Evidence · ANTs

N4BiasFieldCorrection

N4BiasFieldCorrection estimates the smooth bias field that makes one tissue brighter in one part of an image than in another, and divides it out. LARMORX’s port is a replica of ANTs 2.6.5’s program on ITK 5.4.5. In reproduce it gives ANTs’ output bit for bit in every compared case of its two records once ANTs uses correctly rounded maths (46 of 46 and 69 of 69). With ANTs as built, 29 of the 46 and 20 of the 69 match bit for bit; the others differ only where glibc’s maths is not correctly rounded. On fMRIPrep’s own N4 of a real T1w, on any number of threads, it gives ANTs’ one-thread result with correctly rounded maths. Its default, fix, leaves the zero and negative voxels out of the fit (ANTS-40), a correction measured on fMRIPrep’s 25 N4 calls.

Ports
  • replica larmorx Apache-2.0: translated from ANTs 2.6.5's and ITK 5.4.5's source (Apache-2.0)
Status
replica validated, with its data panel (marked before the statuses' definition)
Records
standard comparison (ANTs 2.6.5), data panel (ANTs 2.6.5) · benchmark ants-n4 · findings ants-n4
Reference
N4BiasFieldCorrection · API page ants-n4

What is compared

N4 estimates a smooth multiplicative bias field by fitting B-splines to the residual of a sharpened intensity histogram, level by level, and divides it out. fMRIPrep, sMRIPrep and sdcflows call it on BOLD references, T1w images (before the template and in the brain extraction) and field-map magnitudes.

The replica runs beside ANTs 2.6.5’s N4BiasFieldCorrection on the same files with the same arguments. ANTs is the repository’s build (GCC 14.3), run on one thread as built and with correctly rounded maths, and at more threads for its own variability. fMRIPrep’s own ANTs 2.6.2 runs too, in fMRIPrep’s image.

bash
larmorx ants N4BiasFieldCorrection -d 3 -b '[200]' -c '[50x50x50x50x50,1e-7]' -s 4 -i t1w.nii.gz -o '[t1w_n4.nii.gz,bias.nii.gz]'
python
import larmorx as lx

res = lx.ants.n4_bias_field_correction(
    "sub-01_T1w.nii.gz", shrink_factor=4,
    convergence=([50] * 5, 1e-7), bspline_fitting=200, n_threads=8,
)
corrected, bias = res.corrected, res.bias_field

The API page maps every option. How to read what follows: how to read the evidence.

Its port

ANTs and ITK come under the Apache-2.0 licence, as LARMORX does, so the replica is in larmorx itself. It is translated from ANTs’ and ITK’s source, step by step, to give their results bit for bit. A clean room of ANTs’ programs is planned after FreeSurfer (roadmap).

ANTs’ N4 changes with its thread count; the replica gives ANTs’ single-thread result on any number of threads. It computes the C library’s maths (expf, logf, exp, sincos) correctly rounded, the same on every platform, which is why ANTs as built differs from it in the last bits where glibc’s are not.

The verdicts at a glance

Agreement, replica against the original: ANTs N4BiasFieldCorrection 2.6.5 (standard comparison and data panel, 122 cases of the two records), by category and verdict
fMRIPrep's calls11
options22
inputs13
errors6
data panel: fMRIPrep's calls64
data panel: other images6

Select a segment to list its cases below.

In reproduce, every case that produces an image gives ANTs’ output bit for bit: as ANTs is built (glibc’s maths, one thread), or, where glibc’s expf, logf, exp or sincos is not correctly rounded for a value the run meets, once the same ANTs binary uses CORE-MATH’s correctly rounded functions. The invalid inputs are refused by both programs. One case, inputs/zero-background-known, checks ANTS-40’s correction against ANTs’ own field with a mask of the positive voxels (its verdict is the correction). Nothing is open.

In fMRIPrep: the final N4 of a real T1w

The image viewer below holds the same N4 run by fMRIPrep 25.2.6, by its own ANTs at one thread, by the repository’s ANTs 2.6.5 (as built, with correctly rounded maths, at 1 to 12 threads) and by the replica (1 and 12 threads, reproduce and fix). The replica’s panels in reproduce give ANTs’ result. In fix, the default, ANTS-40’s correction leaves out the few zero voxels this call’s weight image includes (the row replica, fix − reproduce). Choose the panels, the plane, the slice and the window; select a point to read every panel’s exact value there; set two images side by side, swipe or blink between them.

Worked example

fMRIPrep's final N4 of a real T1w

What this shows: In reproduce, on 1 and 12 threads, the replica gives ANTs’ output at one thread with correctly rounded maths, bit for bit; ANTs as built differs from it only where glibc’s maths is not correctly rounded (largest difference 0.031). fMRIPrep’s output, made by ANTs at 8 threads, differs from both as ANTs differs from itself between thread counts. The default, fix, leaves out the few zero voxels this call’s weight image includes (ANTS-40).

sMRIPrep’s last N4 of its brain extraction, weighted by the white-matter posterior, as fMRIPrep 25.2.6 ran it on OpenNeuro ds000005 sub-01: the same arguments and the same input files, read from the fMRIPrep run’s working directory, given to fMRIPrep’s own ANTs, to the repository’s ANTs 2.6.5 and to the replica.

Data
OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01’s T1w (PDDL-1.0); shown inside fMRIPrep’s brain mask, dilated by 2 voxels
Node
fMRIPrep 25.2.6, anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final (sMRIPrep’s final N4), run with NSLOTS=8
Command
N4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image …/anat_template_wf/denoise/mapflow/_denoise0/sub-01_T1w_noise_corrected.nii.gz --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-01_T1w_noise_corrected_corrected.nii.gz, sub-01_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image …/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz
Programs
ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; ANTs 2.6.2.dev1-gff61a3d (conda-forge) in fMRIPrep’s image; CORE-MATH’s correctly rounded functions preloaded where stated; the replica at commit e22378b480f4
Grid
160 × 192 × 192 voxels of 1 × 1.33 × 1.33 mm; shown: 151 × 125 × 111 around the brain
Inputds000005 sub-01 T1was fMRIPrep passed it, after its DenoiseImage
OriginalANTs N4BiasFieldCorrection 2.6.51 thread, as built · the repository's build (GCC 14.3), glibc's maths
fMRIPrep's outputfMRIPrep's outputfMRIPrep 25.2.6, as it ran: ANTs 2.6.2 in its image, 8 threads (NSLOTS=8), glibc's maths
Replicareplicalarmorx, Apache-2.0 · 1 thread, reproduce · against ANTs N4BiasFieldCorrection 2.6.5
the original against itself: 1 and 8 threadsANTs N4BiasFieldCorrection 2.6.5, 8 threads − 1 threadscale ±134
the replica against the originalreplica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtscale ±134

Bit-identical5,898,240 of 5,898,240 voxels equal

the replica against the original: correctly rounded mathsreplica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, correctly rounded maths
the replica against itself: fix and reproduce (ANTS-40)replica, fix − reproduce (ANTS-40)scale ±134

Axial slice 67 of 111 of the shown grid (the one with the most brain), the crosshair where the replica and ANTs as built differ most in it; the corrected image, in the brain's 0.5–99.5 % window; the differences on one symmetric log scale over six decades below ±134, any non-zero difference tinted. The interactive viewer needs JavaScript; the metrics below are complete without it.

What separates fMRIPrep's output from the replica's (the corrected image)

    • fMRIPrep 25.2.6, as it ranANTs 2.6.2 in its image, 8 threads (NSLOTS=8), glibc's maths
    • ANTs N4BiasFieldCorrection 2.6.5, 8 threadsthe repository's build, glibc's maths
    • ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtthe repository's build (GCC 14.3), glibc's maths
    • ANTs N4BiasFieldCorrection 2.6.2, fMRIPrep's build, 1 threadconda-forge's build, glibc's maths
    • ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths, 1 threadthe same binary, CORE-MATH preloaded
    • ANTs N4BiasFieldCorrection 2.6.2, fMRIPrep's build, correctly rounded maths, 1 threadCORE-MATH preloaded
    • replica, reproduce1 and 12 threads

Each box holds runs that give the same bytes, both outputs. fMRIPrep’s ANTs 2.6.2 and the repository’s ANTs 2.6.5 agree bit for bit at equal thread counts and maths, so the version plays no part; ANTs’ thread count and glibc’s maths are the only two causes, and the replica’s reproduce computes ANTs’ result at one thread with correctly rounded maths, on any number of threads. Its default, fix, then corrects ANTS-40 (the row replica, fix − reproduce).

Every comparison, computed in float64 over every voxel of the outputs

Comparison (A − B), verdict and reasonOutputmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
The original against itself: the yardstick
ANTs N4BiasFieldCorrection 2.6.5, 8 threads − 1 threadreferenceANTs’ result moves with its thread count: ITK’s B-spline fit sums per-thread lattices. ANTS-37Corrected image0.06575,317,238 of 5,898,240 (90.1 %)0.0009912.44e−55.26e−12
Bias field1.07e−55,829,414 of 5,898,240 (98.8 %)9.68e−71.35e−53.91e−12
ANTs N4BiasFieldCorrection 2.6.2 − 2.6.5, 1 thread each, as builtreferenceANTs’ version and build play no part at equal threads and maths.both outputsidentical: 0 of 5,898,240 values differ
ANTs N4BiasFieldCorrection 2.6.2 − 2.6.5, 1 thread each, correctly roundedreferenceANTs’ version and build play no part at equal threads and maths.both outputsidentical: 0 of 5,898,240 values differ
ANTs N4BiasFieldCorrection 2.6.5, 2 threads − 1 threadreferenceANTs’ result moves with its thread count: ITK’s B-spline fit sums per-thread lattices. ANTS-37Corrected image0.01985,247,886 of 5,898,240 (89 %)0.0005167.33e−61.54e−12
Bias field4.86e−65,775,806 of 5,898,240 (97.9 %)9.55e−76.13e−63.17e−12
ANTs N4BiasFieldCorrection 2.6.5, 3 threads − 1 threadreferenceANTs’ result moves with its thread count: ITK’s B-spline fit sums per-thread lattices. ANTS-37Corrected image0.02555,205,861 of 5,898,240 (88.3 %)0.001669.46e−62.61e−12
Bias field6.47e−65,725,543 of 5,898,240 (97.1 %)9.42e−78.16e−62.88e−12
ANTs N4BiasFieldCorrection 2.6.5, 4 threads − 1 threadreferenceANTs’ result moves with its thread count: ITK’s B-spline fit sums per-thread lattices. ANTS-37Corrected image0.04135,255,704 of 5,898,240 (89.1 %)0.0006871.53e−52.79e−12
Bias field6.71e−65,771,717 of 5,898,240 (97.9 %)8.85e−78.47e−62.77e−12
ANTs N4BiasFieldCorrection 2.6.5, 12 threads − 1 threadreferenceANTs’ result moves with its thread count: ITK’s B-spline fit sums per-thread lattices. ANTS-37Corrected image0.01755,331,127 of 5,898,240 (90.4 %)0.001166.47e−61.54e−12
Bias field4.68e−65,844,167 of 5,898,240 (99.1 %)1.07e−65.91e−62.36e−12
fMRIPrep's output − ANTs N4BiasFieldCorrection 2.6.5, 8 threadsreferenceANTs’ version and build play no part at equal threads and maths.both outputsidentical: 0 of 5,898,240 values differ
The replica against the original and fMRIPrep's output
replica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtIdentical with correctly rounded mathsOnly where glibc’s expf, logf, exp or sincos is not correctly rounded: with CORE-MATH preloaded, the same ANTs binary gives the replica’s output bit for bit (the row against correctly rounded maths). platform mathsCorrected image0.03125,327,635 of 5,898,240 (90.3 %)0.001811.16e−53.27e−12
Bias field5.78e−65,842,682 of 5,898,240 (99.1 %)1.06e−67.3e−64.38e−12
replica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, correctly rounded mathsBit-identicalWith correctly rounded maths, ANTs computes exactly what the replica’s reproduce computes.both outputsidentical: 0 of 5,898,240 values differ
replica − fMRIPrep's outputWithin the original's own variabilityfMRIPrep’s output is ANTs at 8 threads (bit-identical to ANTs 2.6.5 at 8 threads), the replica’s is ANTs at 1 thread with correctly rounded maths: this is how two runs of ANTs itself differ (its thread count, glibc’s maths). ANTS-37Corrected image0.09675,240,769 of 5,898,240 (88.9 %)0.00253.58e−51.25e−11
Bias field1.56e−55,748,094 of 5,898,240 (97.5 %)1.21e−61.98e−59.51e−12
The replica against itself: thread counts, fix and reproduce
replica, 12 threads − 1 threadBit-identicalThe replica gives ANTs’ single-thread result on any number of threads.both outputsidentical: 0 of 5,898,240 values differ
replica, fix − reproduce (ANTS-40)Deliberate correctionANTS-40, corrected by default with a warning: of the voxels N4 uses (the weight is positive there), those whose value is 0 have no logarithm; ITK keeps them in its log image as the log of an intensity of 1, inside the histogram and the fit. The replica leaves them out, as a mask of the positive voxels would. In reproduce, the replica gives ANTs’ result. ANTS-40Corrected image1435,380,307 of 5,898,240 (91.2 %)13.80.05318.97e−5
Bias field0.05455,898,240 of 5,898,240 (100 %)0.009310.06880.000135
replica, fix, 12 threads − 1 threadBit-identicalThe replica’s default, fix, too gives one result on any number of threads.both outputsidentical: 0 of 5,898,240 values differ

Δ is A − B. Every number is computed in float64 over all 5,898,240 voxels of the 160 × 192 × 192 outputs, not only the brain shown; the range is B’s maximum minus its minimum; NCC is Pearson’s correlation. The record’s thresholds: values bit-identical to ANTs at one thread with correctly rounded math (CORE-MATH preloaded); with ANTs as built (glibc’s math), bit-identity reported and NCC > 0.995 where they differ. The rows the original against itself are ANTs’ own variability on this input, the yardstick for the rest.

Provenance of this example
  • Inputs: fMRIPrep 25.2.6’s run ds000005-sub-01-noreconall (image sha256 7f84379b40f39987…): input (ec055df157667867…); weight image (18144dfca7a49ddd…); fMRIPrep’s corrected image (fe9f33bd9791dfaf…); fMRIPrep’s bias field (932355402736757a…)
  • Runs: figures cache entry ants-n4/fmriprep-ds000005-final-n4, computed 2026-10-21T09:26:11+00:00; 4 of its 13 runs made again since, last on 2026-10-23T22:00:37+00:00, with larmorx e22378b480f4; every command, its environment, threads, exit status and time are in its RUN_INFO.json
  • Data: OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01, PDDL-1.0

Every number above is in ants-n4.json, exported by python -m larmorx_validation evidence ants-n4.

A deliberate correction: zero and negative voxels (ANTS-40)

ITK’s N4 takes the logarithm of the voxels it uses only where they are above 0: a 0 stays 0, the logarithm of an intensity of 1, and a negative value stays negative, far below every tissue’s logarithm. Both enter the histogram N4 sharpens and the B-spline fit. fMRIPrep runs every N4 without a mask, so a zero background, zero-filled voxels and negative values are fitted as if they were tissue. The project’s rules put such a defect to measurement first: every N4 call of the five fMRIPrep runs, run as ANTs does and with those voxels left out (what a mask of the positive voxels does, which a unit test of the replica shows to be the correction itself), against ANTs’ own variability for the same call, and phantoms with a known bias field.

ANTS-40 on fMRIPrep’s 25 N4 calls: the voxels N4 uses that are 0 or negative, and how leaving them out changes the bias field (its global factor; its shape, the ratio over its median, over fMRIPrep’s brain mask, the head for the field maps), beside ANTs’ own variability (1 against 6 threads) and, for the final N4s, the white matter’s CV over the N4’s weight region

Call0 or negativevoxels usedglobal factorshape change, 99th pctshape change, maxANTs' thread band, 99th pctWM CV: inputANTscorrected
ds000005-sub-01: T1w, the brain extraction’s final N4weight image; over the brain810,6570.9550.03040.1671.11e−50.08790.04120.0411
ds000005-sub-01: T1w, the brain extraction’s first N4over the brain092,1601001.99e−5–––
ds000005-sub-01: BOLD reference (task mixedgamblestask run 01)weight image; over the brain616880.9630.1080.2072.99e−6–––
ds000005-sub-01: BOLD reference (task mixedgamblestask run 02)weight image; over the brain06931000.000308–––
ds000005-sub-01: BOLD reference (task mixedgamblestask run 03)weight image; over the brain07041000.0031–––
ds001600-sub-1: T1w, the brain extraction’s final N4weight image; over the brain119,0280.970.005990.02513.34e−60.06960.03580.0357
ds001600-sub-1: T1w, the brain extraction’s first N4over the brain0180,2241002.43e−6–––
ds001600-sub-1: field-map magnitude (wf auto 00000)over the head1,3022,8160.9620.3590.6257.98e−6–––
ds001600-sub-1: field-map magnitude (wf auto 00001)over the head1,3522,8160.9870.09730.4331.49e−5–––
ds001600-sub-1: field-map magnitude (wf auto 00002)over the head1,3682,8160.9990.1930.5134.43e−6–––
ds001600-sub-1: field-map magnitude (wf auto 00003)over the head2,9506,0001.070.1730.3783.62e−6–––
ds005040-sub-002: T1w, the brain extraction’s final N4weight image; over the brain935,6920.7670.05420.114.9e−60.09320.0320.0313
ds005040-sub-002: T1w, the brain extraction’s first N4over the brain0248,8321008.07e−6–––
ds005040-sub-002: BOLD reference (task clip run 1)weight image; over the brain01,0421000.000371–––
ds005040-sub-002: field-map magnitude (wf auto 00000)over the head9812,5601.050.2651.035.53e−6–––
ds005454-sub-16: T1w template, before merging (image 1)over the brain116,646274,1760.8320.07280.111.68e−5–––
ds005454-sub-16: T1w template, before merging (image 2)over the brain81,506274,1761.460.1160.131.74e−5–––
ds005454-sub-16: T1w, the brain extraction’s final N4weight image; over the brain4333,4710.870.3010.3961.11e−50.07620.07620.0689
ds005454-sub-16: T1w, the brain extraction’s first N4over the brain0274,1761001.67e−5–––
ds005454-sub-16: BOLD reference (task rest)weight image; over the brain09,7751002.14e−6–––
ds005454-sub-16: field-map magnitude (wf gre16)over the head6,19013,3280.9780.180.41.04e−5–––
ds005752-sub-ON01016: T1w, the brain extraction’s final N4weight image; over the brain415,2750.9330.01640.03766.87e−60.08030.04980.0497
ds005752-sub-ON01016: T1w, the brain extraction’s first N4over the brain0200,7041001.11e−5–––
ds005752-sub-ON01016: BOLD reference (ses 01 task rest dir forward)weight image; over the brain17781.010.01180.02940.000614–––
ds005752-sub-ON01016: BOLD reference (ses 01 task rest dir reverse)weight image; over the brain37710.9980.02530.05860.00299–––

From validation/investigations/ants-40-n4-nonpositive/analysis.json (measure.py and analyse.py, the replica’s N4 as ANTs runs it with and without a mask of the positive voxels, which the unit test and the comparison’s case inputs/zero-background-known show to be the correction).

Sixteen of the 25 calls use such voxels. Leaving them out changes the field far beyond ANTs’ own variability, even for a handful of voxels among tens of thousands; on the phantoms it recovers the known field better, and on fMRIPrep’s final N4s the white matter becomes more uniform in all five runs.

ANTS-40 on phantoms with a known bias field (±25 %): the error of the estimated field’s shape and the white matter’s CV, ANTs’ way and with the zero voxels left out

Phantomzero voxelsfield error: ANTscorrectedWM CV: inputANTscorrected
zero background768,6400.0260.009670.0660.02710.0255
zero dropout773,9010.02640.009610.1570.1450.145
noisy background00.1110.1110.0660.02570.0257

From validation/investigations/ants-40-n4-nonpositive/phantom.json (phantom.py); the noisy background has no zero voxel, so both runs give the same bits there.

The decision was to correct it by default: a value without a logarithm has no place in N4’s model of the log intensities. The example is the run where it matters most: sMRIPrep’s final N4 of a 7 T T1w template (ds005454 sub-16), which has negative values. Their logarithms stretch the histogram until the tissue sits in one or two of its bins, the sharpening changes nothing, and ANTs’ bias field is flat: it corrects nothing. With them left out, the replica’s fix corrects the image; reproduce gives ANTs’ output.

Worked example

ANTS-40: fMRIPrep's final N4 of a 7T T1w, which ANTs leaves uncorrected

What this shows: ANTs’ bias field on this 7 T template is flat, 1.0000 to 1.0002: the 43 negative voxels N4 uses stretch its histogram, and nothing is corrected. The replica’s default, fix, leaves them out (ANTS-40), and its field spans 0.485 to 1.157; in reproduce it gives ANTs’ output bit for bit with correctly rounded maths.

sMRIPrep’s final N4 of its brain extraction, as fMRIPrep 25.2.6 ran it on OpenNeuro ds005454 sub-16 (a 7T T1w template with negative values), its arguments and inputs from the fMRIPrep run: a few of the voxels N4 uses are negative, ITK keeps them in its log image, and ANTs’ field stays flat. The replica’s default, fix, leaves them out of the fit and corrects the image; reproduce gives ANTs’ result.

Data
OpenNeuro ds005454 (fMRIPrep boot camp, 7T training dataset), v1.0.0, sub-16 (CC-BY-4.0)
Node
fMRIPrep 25.2.6, anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final
Command
N4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image …/anat_fit_wf/anat_template_wf/anat_merge/sub-16_acq-denoised_T1w_noise_corrected_template.nii.gz --convergence [ 50x50x50x50x50, 1e-07 ] --output [corrected.nii.gz,bias.nii.gz] -r --shrink-factor 4 --weight-image …/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz
Defect
ANTS-40: N4 keeps zero and negative values in the log image; corrected by default (decided 2026-10-23), reproduce gives ANTs’ result
Effect
43 of the 33,471 voxels N4 uses are 0 or negative (0 zero, 43 negative; the replica’s notice); ANTs’ bias field spans 1.0000 to 1.0002, the replica’s 0.485 to 1.157; Corrected image: 17,379,407 of 17,547,264 values change, by at most 2.03e+03; Bias field: 17,547,263 of 17,547,264 values change, by at most 0.515
Programs
ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0, one thread; the replica at commit e22378b480f4, 6 threads
Grid
224 × 272 × 288 voxels of 0.8 × 0.799 × 0.799 mm; shown: 172 × 214 × 172 around the brain

Data shown

Inputds005454 sub-16 T1w template (7T)as fMRIPrep passed it, after DenoiseImage
OriginalANTs N4BiasFieldCorrection 2.6.51 thread, as built · the repository's build (GCC 14.3), glibc's maths
Replicareplicalarmorx, Apache-2.0 · fix (the default): ANTS-40 corrected
Replicareplicalarmorx, Apache-2.0 · reproduce: ANTs' behaviour · against ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths

Bit-identical17,547,264 of 17,547,264 voxels equal

the replica against the original: correctly rounded mathsreplica, reproduce − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths
the replica against itself: fix and reproduce (ANTS-40)replica, fix − reproduce (ANTS-40)scale ±1.35e3

Axial slice 109 of 172; the crosshair where the replica's correction changes the output most in it. The interactive viewer needs JavaScript; the metrics below are complete without it.

Every comparison, computed in float64 over every voxel of the outputs

Comparison (A − B), verdict and reasonOutputmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
The original against itself: the yardstick
ANTs N4BiasFieldCorrection 2.6.5: correctly rounded maths − glibc'sreferenceWhere glibc’s exp, log, sin or cos is not correctly rounded for a value the run meets: ANTs’ own result moves with the maths library.Corrected image0.03312,470,213 of 17,547,264 (71.1 %)0.002161.12e−55.57e−12
Bias field1.45e−512,476,998 of 17,547,264 (71.1 %)1.8e−60.1230.00274
fMRIPrep's output − ANTs N4BiasFieldCorrection 2.6.5, 1 threadreferencefMRIPrep’s run used its own ANTs build and thread count: ANTs’ own variability.Corrected image0.033212,160,508 of 17,547,264 (69.3 %)0.002121.13e−55.37e−12
Bias field1.45e−512,169,422 of 17,547,264 (69.4 %)1.77e−60.1230.00264
The replica against the original
replica, fix − ANTs N4BiasFieldCorrection 2.6.5, as builtDeliberate correctionANTS-40, corrected by default with a warning: ITK keeps the voxels N4 uses whose value is 0 or negative in its log image as they are; here negative values, whose logs lie far below every tissue’s, stretch the histogram until the tissue sits in one or two of its bins, the sharpening changes nothing and ANTs’ field stays flat. The replica leaves them out of the fit, as a mask of the positive voxels would; reproduce gives ANTs’ result. ANTS-40Corrected image2.03e317,379,407 of 17,547,264 (99 %)3010.690.0222
Bias field0.51517,547,263 of 17,547,264 (100 %)0.2094.37e31.56
replica, reproduce − ANTs N4BiasFieldCorrection 2.6.5, as builtIdentical with correctly rounded mathsOnly where glibc’s maths is not correctly rounded: with CORE-MATH preloaded, ANTs gives the replica’s output in reproduce bit for bit (the row against correctly rounded maths).Corrected image0.03312,470,213 of 17,547,264 (71.1 %)0.002161.12e−55.57e−12
Bias field1.45e−512,476,998 of 17,547,264 (71.1 %)1.8e−60.1230.00274
replica, reproduce − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded mathsBit-identicalIn reproduce, the replica gives ANTs’ result, the defect included.both outputsidentical: 0 of 17,547,264 values differ
The replica against itself: fix and reproduce
replica, fix − reproduce (ANTS-40)Deliberate correctionANTS-40, corrected by default with a warning: ITK keeps the voxels N4 uses whose value is 0 or negative in its log image as they are; here negative values, whose logs lie far below every tissue’s, stretch the histogram until the tissue sits in one or two of its bins, the sharpening changes nothing and ANTs’ field stays flat. The replica leaves them out of the fit, as a mask of the positive voxels would; reproduce gives ANTs’ result. ANTS-40Corrected image2.03e317,379,407 of 17,547,264 (99 %)3010.690.0222
Bias field0.51517,547,263 of 17,547,264 (100 %)0.2094.51e31.55

Δ is A − B, over all 17,547,264 voxels of each output file; the rows the original against itself are ANTs’ own variability on this input.

Provenance of this example
  • Data: OpenNeuro ds005454 (fMRIPrep boot camp, 7T training dataset), v1.0.0, sub-16, CC-BY-4.0
  • Inputs: input (sha256 065c3abba40a2cb5…); weight image (sha256 c8081b480c163d71…); fMRIPrep’s corrected (sha256 96eac1e300adfc83…); fMRIPrep’s bias (sha256 85a1f00e5a39fd93…)
  • Runs: figures cache entry ants-n4/fmriprep-ds005454-final-n4-ants40, computed 2026-10-23T22:04:37+00:00 with larmorx e22378b480f4; ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; every command, its environment, threads and exit status are in its RUN_INFO.json

Every number above is in ants-n4.json, exported by python -m larmorx_validation evidence ants-n4.

Against the original’s own variability

ANTs’ N4 changes with its thread count (ANTS-37): ITK’s B-spline fit sums per-thread lattices. The record measures that band for every case that differs from ANTs as built, and divides the replica’s difference by it.

Variability, the replica against the original divided by the original against itself between thread counts: ANTs N4BiasFieldCorrection 2.6.5, the 17 cases that differ from it as built (log scale)
  • the replica's largest |Δ| from ANTs ÷ ANTs' largest |Δ| between its thread counts
  • the replica's (1 − NCC) with ANTs ÷ ANTs' (1 − NCC) between its thread counts
  • inside the original's own variability, ANTs between its thread counts (< 1)
  1. fmriprep/boldref-coregistration correctedlargest |Δ| 0.43 × the band1 − NCC 0.45 × the band
  2. fmriprep/boldref-template correctedlargest |Δ| 0.25 × the band1 − NCC 0 × the band
  3. fmriprep/smriprep-pre-template correctedlargest |Δ| 1.3 × the band1 − NCC 0.24 × the band
  4. fmriprep/n4-only correctedlargest |Δ| 0.25 × the band1 − NCC 0.01 × the band
  5. fmriprep/n4-only bias fieldlargest |Δ| 0.15 × the band1 − NCC 0.0053 × the band
  6. fmriprep/t1w-ds000210-final correctedlargest |Δ| 0.21 × the band1 − NCC 0.064 × the band
  7. fmriprep/t1w-ds000210-final bias fieldlargest |Δ| 0.21 × the band1 − NCC 0.17 × the band
  8. fmriprep/t1w-preproc-mask correctedlargest |Δ| 0.023 × the band1 − NCC 0.015 × the band
  9. options/mask-and-weight correctedlargest |Δ| 0.87 × the band1 − NCC 0.79 × the band
  10. options/shrink-1 correctedlargest |Δ| 0.36 × the band1 − NCC 0.26 × the band
  11. options/shrink-3 correctedlargest |Δ| 1.7 × the band1 − NCC 4.2 × the band
  12. options/shrink-0 correctedlargest |Δ| 0.36 × the band1 − NCC 0.26 × the band
  13. options/distance-30 correctedlargest |Δ| 0.4 × the band1 − NCC 0.058 × the band
  14. options/distance-30 bias fieldlargest |Δ| 0.4 × the band1 − NCC 0.092 × the band
  15. options/distance-order-2 correctedlargest |Δ| 0.22 × the band1 − NCC 0.0028 × the band
  16. inputs/int16-scaled correctedlargest |Δ| 0.072 × the band1 − NCC 0.0031 × the band
  17. inputs/zero-background-known correctedlargest |Δ| 2.2 × the band1 − NCC 5.3 × the band
  18. inputs/zero-background-known bias fieldlargest |Δ| 2.4 × the band1 − NCC 4.7 × the band
  19. inputs/las-sform-only correctedlargest |Δ| 0.46 × the band1 − NCC 0.31 × the band
  20. inputs/dim2-distance correctedlargest |Δ| 0.33 × the band1 − NCC 0.56 × the band
  21. inputs/bold-first-volume correctedlargest |Δ| 0.39 × the band1 − NCC 0.73 × the band
0.001×0.01×0.1×1×10×

From the record’s table: ANTs at 2, 3, 4, 8 and 12 threads against 1 thread give ANTs’ thread band for each case and output; a ratio below 1 lies inside it. Every one of these differences vanishes with correctly rounded maths.

On the data panel

The data panel is the fixed set of fourteen OpenNeuro subjects on which LARMORX’s programs are validated: 1.5, 3 and 7 T; Siemens, GE and Philips; single-band, multiband and multi-echo BOLD; a 3.5-year-old to an 82-year-old; a tumour, a stroke with its lesion mask, strong motion. Its record replays every call fMRIPrep 25.2.6 made of the program in its fourteen runs on these subjects, the node’s own command.txt with its inputs from the run’s working directory (calls repeated word for word are compared once), and runs the program on panel images fMRIPrep never gives it. ANTs runs at one thread, as built and with correctly rounded maths, and at 2 and 3 threads for its own variability where it has any; the replica in reproduce, and in fix for the corrections. The panel’s cases are in the case explorer below (the categories data panel: fMRIPrep’s calls and data panel: other images).

69 of 69 bit-identical to ANTs with correctly rounded maths, 20 to ANTs as built: all 64 N4 calls of the fourteen runs (the BOLD references, the brain extraction’s and the final anatomical corrections, the field maps’ magnitudes) and sMRIPrep’s final options on five images fMRIPrep never corrects (the 7 T MP2RAGE, a 3 T MP2RAGE UNIT1, the 82-year-old’s FLAIR, a glioma’s T2w of 6 mm slices, the stroke’s T2w fitted outside its lesion). Against ANTs as built every difference is within 0.042 of ANTs’ own difference between one thread and two or three. A mask written from the image’s affine (2e-6 mm off its grid) is refused by both, as ITK’s input check requires.

fMRIPrep’s N4BiasFieldCorrection nodes in the fourteen fMRIPrep runs, and what the panel record compares of them

Node (normalised path)NodesRunsComparedSame call as a compared oneStatusesNot compared: why
anat_fit_wf/anat_template_wf/n4_correct52505 pass–
anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final141414014 pass–
anat_fit_wf/brain_extraction_wf/inu_n4141414014 pass–
bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct221222022 pass–
fmap_preproc_wf/wf_auto_00000/brainextraction_wf/n411101 pass–
fmap_preproc_wf/wf_auto_00000/magnitude_wf/brainextraction_wf/n433303 pass–
fmap_preproc_wf/wf_auto_00001/magnitude_wf/brainextraction_wf/n422202 pass–
fmap_preproc_wf/wf_auto_00002/magnitude_wf/brainextraction_wf/n411101 pass–
fmap_preproc_wf/wf_auto_00003/brainextraction_wf/n411101 pass–
fmap_preproc_wf/wf_gre16/magnitude_wf/brainextraction_wf/n411101 pass–

From the panel record (docs/validation/ants-n4-panel.md): every node of the finished fMRIPrep 25.2.6 runs that ran the program, grouped by its path with the subject, run, space and item removed; a call repeated word for word is compared once.

Images on a small grid (a BOLD reference, a template at 2 mm) are shown as whole volumes; the panel’s large anatomical images (up to 17.5 million voxels) as three full-resolution slices through the voxel where the replica and ANTs as built differ most (the centre of the field of view where they agree), the whole field of view shown, defaced heads included (the data panel’s defacing check).

Worked example

Data panel: fMRIPrep's N4 of a multiband-8 BOLD reference

What this shows: In reproduce, on 1 and 8 threads, the replica gives ANTs’ bytes, as built and with correctly rounded maths. No correction by default changes this call.

fMRIPrep 25.2.6’s own call (the BOLD reference’s correction before its skull stripping, n4_correct: -b [200], -r without a mask, its weight image) on ds004958 sub-fsm64ps’s multiband-8 run (600 volumes of 104 × 104 × 72 at 2 mm): its command.txt and inputs from the fMRIPrep run. A key dataset, shown as a whole volume.

Data
OpenNeuro ds004958, sub-fsm64ps (CC0-1.0); the whole field of view shown, only margins where every array is 0 left out
Node
bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correct of the fMRIPrep run ds004958-sub-fsm64ps-noreconall
Command
N4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image sub-fsm64ps_task-rest_sbref.nii.gz --output sub-fsm64ps_task-rest_sbref_corrected.nii.gz -r --weight-image tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz
Programs
ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; the replica at commit 042d70d5a17c
Grid
104 × 104 × 72 voxels of 2 × 2 × 2 mm; shown 104 × 104 × 72, every slice

Data shown

Inputds004958 sub-fsm64ps, BOLD reference (fMRIPrep's)sub-fsm64ps_task-rest_sbref.nii.gz
OriginalANTs N4BiasFieldCorrection 2.6.51 thread · as built (GCC 14.3), glibc's maths
Replicareplicalarmorx, Apache-2.0 · reproduce, 8 threads · against ANTs N4BiasFieldCorrection 2.6.5
Replicareplicalarmorx, Apache-2.0 · fix (the default), 8 threads

Bit-identical778,752 of 778,752 voxels equal

the replica against the originalreplica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as built

Bit-identical778,752 of 778,752 voxels equal

the replica against the original: correctly rounded mathsreplica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths

Bit-identical778,752 of 778,752 voxels equal

the replica against itself: fix and reproducereplica, fix − reproduce

Axial slice 37 of 72; the crosshair at the centre of the field of view. The interactive viewer needs JavaScript; the metrics below are complete without it.

Every comparison, computed in float64 over every voxel of the outputs

Comparison (A − B), verdict and reasonmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
The original against itself: the yardstick
ANTs N4BiasFieldCorrection 2.6.5, 8 threads − 1 threadreferenceANTs’ own result moves with its thread count (corrected): the yardstick for any difference above.24.4766,603 of 778,752 (98.4 %)0.6760.002126.53e−8
The replica against the original
replica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtBit-identicalThe replica gives the unmodified ANTs program’s bytes.identical: 0 of 778,752 values differ
replica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded mathsBit-identicalThe replica gives ANTs’ bytes with correctly rounded maths.identical: 0 of 778,752 values differ
The replica against itself: thread counts, fix and reproduce
replica, fix − reproduceBit-identicalNo correction of the defect register changes this call.identical: 0 of 778,752 values differ
replica, 1 thread − 8 threads (reproduce)Bit-identicalThe replica’s result does not depend on its thread count.identical: 0 of 778,752 values differ

Δ is A − B over all 778,752 values of each output.

Provenance of this example
  • Data: OpenNeuro ds004958, sub-fsm64ps, CC0-1.0
  • Inputs: input 1 (sha256 1258741f225e0f58…)
  • Runs: figures cache entry ants-n4/panel-fmriprep-boldref-mb8, computed 2026-10-27T01:53:24+00:00 with larmorx 042d70d5a17c; ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; every command, its environment, threads and exit status are in its RUN_INFO.json

Every number above is in ants-n4.json, exported by python -m larmorx_validation evidence ants-n4.

Worked example

Data panel: sMRIPrep's final N4 options on the raw 7 T MP2RAGE

What this shows: In reproduce, on 1 and 8 threads, the replica gives ANTs’ bytes with correctly rounded maths; ANTs as built differs only where glibc’s maths is not correctly rounded (largest difference 0.041). Its default, fix, corrects ANTS-40 here.

N4 with sMRIPrep’s final options (-b [200] -c [50x50x50x50x50,1e-7] -s 4, the bias field saved) on the raw 7 T MP2RAGE uniform image of ds005454 sub-16, the panel’s largest anatomical image: its background noise as bright as the head, fitted with it (no mask).

Data
OpenNeuro ds005454, sub-16 (CC-BY-4.0); the whole field of view shown, only margins where every array is 0 left out
Command
N4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image sub-16_acq-mp2rage_T1w.nii.gz --output [ corrected.nii.gz, bias.nii.gz ]
Programs
ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; the replica at commit 042d70d5a17c
Grid
224 × 272 × 288 voxels of 0.8 × 0.799 × 0.799 mm; shown 224 × 272 × 288, three full-resolution slices

Data shown

Inputds005454 sub-16sub-16_acq-mp2rage_T1w.nii.gz
OriginalANTs N4BiasFieldCorrection 2.6.51 thread · as built (GCC 14.3), glibc's maths
Replicareplicalarmorx, Apache-2.0 · reproduce, 8 threads · against ANTs N4BiasFieldCorrection 2.6.5
Replicareplicalarmorx, Apache-2.0 · fix (the default), 8 threads
the replica against the originalreplica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtscale ±1.14e3

Bit-identical17,547,264 of 17,547,264 voxels equal

the replica against the original: correctly rounded mathsreplica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths
the replica against itself: fix and reproducereplica, fix − reproducescale ±1.14e3

Axial slice 64 of 288; the crosshair where the replica differs most from ANTs as built; the three slices through it are stored at full resolution. The interactive viewer needs JavaScript; the metrics below are complete without it.

Stored for this example: three full-resolution slices, one per plane, through display voxel (111, 117, 64): where the main comparison differs most, or the field's centre where nothing differs; every metric is computed over every voxel of the files.

Every comparison, computed in float64 over every voxel of the outputs

Comparison (A − B), verdict and reasonOutputmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
The original against itself: the yardstick
ANTs N4BiasFieldCorrection 2.6.5, 8 threads − 1 threadreferenceANTs’ own result moves with its thread count (corrected, bias): the yardstick for any difference above.corrected0.14612,404,203 of 17,547,264 (70.7 %)0.04562.05e−59.72e−13
bias field2.22e−517,547,264 of 17,547,264 (100 %)1.45e−53.89e−54.92e−11
The replica against the original
replica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtIdentical with correctly rounded mathsOnly where glibc’s maths is not correctly rounded (corrected, bias): with CORE-MATH preloaded, ANTs gives the replica’s bytes.corrected0.041312,354,443 of 17,547,264 (70.4 %)0.006445.78e−62.04e−12
bias field9.42e−617,503,900 of 17,547,264 (99.8 %)2.83e−61.65e−51.06e−10
replica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded mathsBit-identicalThe replica gives ANTs’ bytes with correctly rounded maths.both outputsidentical: 0 of 17,547,264 values differ
The replica against itself: thread counts, fix and reproduce
replica, fix − reproduceDeliberate correctionThe replica’s default, fix, corrects ANTS-40 (register); reproduce gives ANTs’ result. Changed: corrected, bias.corrected1.35e312,404,182 of 17,547,264 (70.7 %)81.40.1890.000995
bias field0.21617,547,247 of 17,547,264 (100 %)0.04570.380.132
replica, 1 thread − 8 threads (reproduce)Bit-identicalThe replica’s result does not depend on its thread count.both outputsidentical: 0 of 17,547,264 values differ

Δ is A − B over all 17,547,264 values of each output.

Provenance of this example
  • Data: OpenNeuro ds005454, sub-16, CC-BY-4.0
  • Inputs: input 1 (sha256 971678c9fb76ae12…)
  • Runs: figures cache entry ants-n4/panel-final-7t-mp2rage, computed 2026-10-27T01:12:26+00:00 with larmorx 042d70d5a17c; ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; every command, its environment, threads and exit status are in its RUN_INFO.json

Every number above is in ants-n4.json, exported by python -m larmorx_validation evidence ants-n4.

Worked example

Data panel: fMRIPrep's final N4 of a brain with a tumour

What this shows: In reproduce, on 1 and 8 threads, the replica gives ANTs’ bytes with correctly rounded maths; ANTs as built differs only where glibc’s maths is not correctly rounded (largest difference 0.0068). Its default, fix, leaves the zero and negative voxels out of the fit (ANTS-40) and warns that -r does nothing without a mask (ANTS-36).

fMRIPrep 25.2.6’s own call (sMRIPrep’s final N4, inu_n4_final, its white-matter weight image) on ds005003 sub-RS001, a 70 mL right frontal oligodendroglioma: its command.txt and inputs from the fMRIPrep run.

Data
OpenNeuro ds005003, sub-RS001 (CC0-1.0); the whole field of view shown, only margins where every array is 0 left out
Node
anat_fit/brain_extraction/atropos/inu_n4_final[0] of the fMRIPrep run ds005003-sub-RS001-noreconall
Command
N4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image sub-RS001_ses-1_T1w_noise_corrected.nii.gz --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-RS001_ses-1_T1w_noise_corrected_corrected.nii.gz, sub-RS001_ses-1_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image prodmap.nii.gz
Programs
ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; the replica at commit 042d70d5a17c
Grid
144 × 256 × 256 voxels of 1.2 × 0.977 × 0.977 mm; shown 144 × 256 × 256, three full-resolution slices
Inputds005003 sub-RS001, denoised T1w (fMRIPrep's)sub-RS001_ses-1_T1w_noise_corrected.nii.gz
OriginalANTs N4BiasFieldCorrection 2.6.51 thread · as built (GCC 14.3), glibc's maths
Replicareplicalarmorx, Apache-2.0 · reproduce, 8 threads · against ANTs N4BiasFieldCorrection 2.6.5
Replicareplicalarmorx, Apache-2.0 · fix (the default), 8 threads
the replica against the originalreplica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtscale ±170

Bit-identical9,437,184 of 9,437,184 voxels equal

the replica against the original: correctly rounded mathsreplica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths
the replica against itself: fix and reproducereplica, fix − reproducescale ±170

Axial slice 179 of 256; the crosshair where the replica differs most from ANTs as built; the three slices through it are stored at full resolution. The interactive viewer needs JavaScript; the metrics below are complete without it.

Stored for this example: three full-resolution slices, one per plane, through display voxel (23, 145, 179): where the main comparison differs most, or the field's centre where nothing differs; every metric is computed over every voxel of the files.

Every comparison, computed in float64 over every voxel of the outputs

Comparison (A − B), verdict and reasonOutputmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
The original against itself: the yardstick
ANTs N4BiasFieldCorrection 2.6.5, 8 threads − 1 threadreferenceANTs’ own result moves with its thread count (corrected, bias): the yardstick for any difference above.corrected0.01157,691,288 of 9,437,184 (81.5 %)0.001723.93e−62.44e−13
corrected2.32e−69,425,297 of 9,437,184 (99.9 %)1.15e−62.54e−61.5e−12
The replica against the original
replica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtIdentical with correctly rounded mathsOnly where glibc’s maths is not correctly rounded (corrected, bias): with CORE-MATH preloaded, ANTs gives the replica’s bytes.corrected0.006847,487,867 of 9,437,184 (79.3 %)0.00112.34e−69.74e−13
corrected1.4e−69,128,052 of 9,437,184 (96.7 %)4.08e−71.53e−64.82e−13
replica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded mathsBit-identicalThe replica gives ANTs’ bytes with correctly rounded maths.both outputsidentical: 0 of 9,437,184 values differ
The replica against itself: thread counts, fix and reproduce
replica, fix − reproduceDeliberate correctionThe replica’s default, fix, corrects ANTS-40 (register) and warns that -r does nothing without a mask (ANTS-36); reproduce gives ANTs’ result. Changed: corrected, bias.corrected1987,704,112 of 9,437,184 (81.6 %)30.30.06770.000217
corrected0.02899,437,184 of 9,437,184 (100 %)0.01340.03150.000123
replica, 1 thread − 8 threads (reproduce)Bit-identicalThe replica’s result does not depend on its thread count.both outputsidentical: 0 of 9,437,184 values differ

Δ is A − B over all 9,437,184 values of each output.

Provenance of this example
  • Data: OpenNeuro ds005003, sub-RS001, CC0-1.0
  • Inputs: input 1 (sha256 8f1ed49a919fb45f…)
  • Runs: figures cache entry ants-n4/panel-fmriprep-final-tumour, computed 2026-10-27T01:17:36+00:00 with larmorx 042d70d5a17c; ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; every command, its environment, threads and exit status are in its RUN_INFO.json

Every number above is in ants-n4.json, exported by python -m larmorx_validation evidence ants-n4.

Worked example

Data panel: N4 on a FLAIR of 3 mm slices

What this shows: In reproduce, on 1 and 8 threads, the replica gives ANTs’ bytes, as built and with correctly rounded maths. Its default, fix, corrects ANTS-40 here.

The same options on an 82-year-old’s FLAIR (ds003592 sub-280, 0.42 × 0.42 × 3 mm): shrink 4 leaves 11 slices for the fit.

Data
OpenNeuro ds003592, sub-280 (CC0-1.0); the whole field of view shown, only margins where every array is 0 left out
Command
N4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image sub-280_ses-1_FLAIR.nii.gz --output [ corrected.nii.gz, bias.nii.gz ]
Programs
ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; the replica at commit 042d70d5a17c
Grid
384 × 512 × 44 voxels of 0.422 × 0.422 × 3 mm; shown 384 × 512 × 44, three full-resolution slices
Inputds003592 sub-280sub-280_ses-1_FLAIR.nii.gz
OriginalANTs N4BiasFieldCorrection 2.6.51 thread · as built (GCC 14.3), glibc's maths
Replicareplicalarmorx, Apache-2.0 · reproduce, 8 threads · against ANTs N4BiasFieldCorrection 2.6.5
Replicareplicalarmorx, Apache-2.0 · fix (the default), 8 threads

Bit-identical8,650,752 of 8,650,752 voxels equal

the replica against the originalreplica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as built

Bit-identical8,650,752 of 8,650,752 voxels equal

the replica against the original: correctly rounded mathsreplica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths
the replica against itself: fix and reproducereplica, fix − reproducescale ±16.3

Axial slice 22 of 44; the crosshair at the centre of the field of view; the three slices through it are stored at full resolution. The interactive viewer needs JavaScript; the metrics below are complete without it.

Stored for this example: three full-resolution slices, one per plane, through display voxel (192, 256, 22): where the main comparison differs most, or the field's centre where nothing differs; every metric is computed over every voxel of the files.

Every comparison, computed in float64 over every voxel of the outputs

Comparison (A − B), verdict and reasonOutputmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
The original against itself: the yardstick
ANTs N4BiasFieldCorrection 2.6.5, 8 threads − 1 threadreferenceANTs’ own result moves with its thread count (corrected, bias): the yardstick for any difference above.corrected0.1538,427,472 of 8,650,752 (97.4 %)0.001710.0002159e−11
bias field0.0005098,650,137 of 8,650,752 (100 %)3.74e−50.0002283.01e−9
The replica against the original
replica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtBit-identicalThe replica gives the unmodified ANTs program’s bytes.both outputsidentical: 0 of 8,650,752 values differ
replica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded mathsBit-identicalThe replica gives ANTs’ bytes with correctly rounded maths.both outputsidentical: 0 of 8,650,752 values differ
The replica against itself: thread counts, fix and reproduce
replica, fix − reproduceDeliberate correctionThe replica’s default, fix, corrects ANTS-40 (register); reproduce gives ANTs’ result. Changed: corrected, bias.corrected34.38,428,169 of 8,650,752 (97.4 %)1.080.0488.75e−5
bias field0.2198,650,747 of 8,650,752 (100 %)0.02010.09820.0022
replica, 1 thread − 8 threads (reproduce)Bit-identicalThe replica’s result does not depend on its thread count.both outputsidentical: 0 of 8,650,752 values differ

Δ is A − B over all 8,650,752 values of each output.

Provenance of this example
  • Data: OpenNeuro ds003592, sub-280, CC0-1.0
  • Inputs: input 1 (sha256 08c1cf34701df2fd…)
  • Runs: figures cache entry ants-n4/panel-final-flair, computed 2026-10-27T01:26:21+00:00 with larmorx 042d70d5a17c; ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; every command, its environment, threads and exit status are in its RUN_INFO.json

Every number above is in ants-n4.json, exported by python -m larmorx_validation evidence ants-n4.

Worked example

Data panel: N4 on a T2w of 6 mm slices with a tumour

What this shows: In reproduce, on 1 and 8 threads, the replica gives ANTs’ bytes with correctly rounded maths; ANTs as built differs only where glibc’s maths is not correctly rounded (largest difference 0.019). Its default, fix, corrects ANTS-40 here.

The same options on the tumour’s T2w (ds005003 sub-RS001: 5 mm slices 6 mm apart): shrink 4 leaves 7 slices for the fit.

Data
OpenNeuro ds005003, sub-RS001 (CC0-1.0); the whole field of view shown, only margins where every array is 0 left out
Command
N4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image sub-RS001_ses-1_T2w.nii.gz --output [ corrected.nii.gz, bias.nii.gz ]
Programs
ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; the replica at commit 042d70d5a17c
Grid
320 × 320 × 30 voxels of 0.719 × 0.719 × 6 mm; shown 320 × 320 × 30, three full-resolution slices
Inputds005003 sub-RS001sub-RS001_ses-1_T2w.nii.gz
OriginalANTs N4BiasFieldCorrection 2.6.51 thread · as built (GCC 14.3), glibc's maths
Replicareplicalarmorx, Apache-2.0 · reproduce, 8 threads · against ANTs N4BiasFieldCorrection 2.6.5
Replicareplicalarmorx, Apache-2.0 · fix (the default), 8 threads
the replica against the originalreplica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtscale ±139

Bit-identical3,072,000 of 3,072,000 voxels equal

the replica against the original: correctly rounded mathsreplica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths
the replica against itself: fix and reproducereplica, fix − reproducescale ±139

Axial slice 9 of 30; the crosshair where the replica differs most from ANTs as built; the three slices through it are stored at full resolution. The interactive viewer needs JavaScript; the metrics below are complete without it.

Stored for this example: three full-resolution slices, one per plane, through display voxel (210, 277, 9): where the main comparison differs most, or the field's centre where nothing differs; every metric is computed over every voxel of the files.

Every comparison, computed in float64 over every voxel of the outputs

Comparison (A − B), verdict and reasonOutputmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
The original against itself: the yardstick
ANTs N4BiasFieldCorrection 2.6.5, 8 threads − 1 threadreferenceANTs’ own result moves with its thread count (corrected, bias): the yardstick for any difference above.corrected0.004762,374,419 of 3,072,000 (77.3 %)0.0006942.91e−66.7e−14
bias field1.17e−53,072,000 of 3,072,000 (100 %)4.22e−66.43e−64.81e−12
The replica against the original
replica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtIdentical with correctly rounded mathsOnly where glibc’s maths is not correctly rounded (corrected, bias): with CORE-MATH preloaded, ANTs gives the replica’s bytes.corrected0.01922,374,386 of 3,072,000 (77.3 %)0.001541.18e−56.16e−12
bias field7.15e−53,071,975 of 3,072,000 (100 %)1.53e−53.94e−56.02e−10
replica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded mathsBit-identicalThe replica gives ANTs’ bytes with correctly rounded maths.both outputsidentical: 0 of 3,072,000 values differ
The replica against itself: thread counts, fix and reproduce
replica, fix − reproduceDeliberate correctionThe replica’s default, fix, corrects ANTS-40 (register); reproduce gives ANTs’ result. Changed: corrected, bias.corrected2302,374,419 of 3,072,000 (77.3 %)11.90.140.00058
bias field0.6983,072,000 of 3,072,000 (100 %)0.09970.3840.139
replica, 1 thread − 8 threads (reproduce)Bit-identicalThe replica’s result does not depend on its thread count.both outputsidentical: 0 of 3,072,000 values differ

Δ is A − B over all 3,072,000 values of each output.

Provenance of this example
  • Data: OpenNeuro ds005003, sub-RS001, CC0-1.0
  • Inputs: input 1 (sha256 f9d0b881a767a40c…)
  • Runs: figures cache entry ants-n4/panel-final-tumour-t2w, computed 2026-10-27T01:29:29+00:00 with larmorx 042d70d5a17c; ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; every command, its environment, threads and exit status are in its RUN_INFO.json

Every number above is in ants-n4.json, exported by python -m larmorx_validation evidence ants-n4.

Worked example

Data panel: N4 on a 3 T MP2RAGE UNIT1

What this shows: In reproduce, on 1 and 8 threads, the replica gives ANTs’ bytes with correctly rounded maths; ANTs as built differs only where glibc’s maths is not correctly rounded (largest difference 0.015). No correction by default changes this call.

The same options on a 68-year-old’s 3 T MP2RAGE UNIT1 (ds004958 sub-fsm64ps), its salt-and-pepper background fitted with the head.

Data
OpenNeuro ds004958, sub-fsm64ps (CC0-1.0); the whole field of view shown, only margins where every array is 0 left out
Command
N4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image sub-fsm64ps_acq-mp2rage_UNIT1.nii.gz --output [ corrected.nii.gz, bias.nii.gz ]
Programs
ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; the replica at commit 042d70d5a17c
Grid
176 × 240 × 256 voxels of 1 × 1 × 1 mm; shown 176 × 240 × 256, three full-resolution slices
Inputds004958 sub-fsm64pssub-fsm64ps_acq-mp2rage_UNIT1.nii.gz
OriginalANTs N4BiasFieldCorrection 2.6.51 thread · as built (GCC 14.3), glibc's maths
Replicareplicalarmorx, Apache-2.0 · reproduce, 8 threads · against ANTs N4BiasFieldCorrection 2.6.5
Replicareplicalarmorx, Apache-2.0 · fix (the default), 8 threads
the replica against the originalreplica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtscale ±0.0151

Bit-identical10,813,440 of 10,813,440 voxels equal

the replica against the original: correctly rounded mathsreplica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded maths

Bit-identical10,813,440 of 10,813,440 voxels equal

the replica against itself: fix and reproducereplica, fix − reproduce

Axial slice 114 of 256; the crosshair where the replica differs most from ANTs as built; the three slices through it are stored at full resolution. The interactive viewer needs JavaScript; the metrics below are complete without it.

Stored for this example: three full-resolution slices, one per plane, through display voxel (131, 121, 114): where the main comparison differs most, or the field's centre where nothing differs; every metric is computed over every voxel of the files.

Every comparison, computed in float64 over every voxel of the outputs

Comparison (A − B), verdict and reasonOutputmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
The original against itself: the yardstick
ANTs N4BiasFieldCorrection 2.6.5, 8 threads − 1 threadreferenceANTs’ own result moves with its thread count (corrected, bias): the yardstick for any difference above.corrected0.11310,726,320 of 10,813,440 (99.2 %)0.04531.33e−51.02e−11
bias field2.21e−510,813,440 of 10,813,440 (100 %)9.7e−62.46e−52.41e−11
The replica against the original
replica − ANTs N4BiasFieldCorrection 2.6.5, 1 thread, as builtIdentical with correctly rounded mathsOnly where glibc’s maths is not correctly rounded (corrected, bias): with CORE-MATH preloaded, ANTs gives the replica’s bytes.corrected0.015110,713,781 of 10,813,440 (99.1 %)0.006381.77e−62.12e−12
bias field4.05e−610,803,732 of 10,813,440 (99.9 %)1.34e−64.52e−64.81e−12
replica − ANTs N4BiasFieldCorrection 2.6.5, correctly rounded mathsBit-identicalThe replica gives ANTs’ bytes with correctly rounded maths.both outputsidentical: 0 of 10,813,440 values differ
The replica against itself: thread counts, fix and reproduce
replica, fix − reproduceBit-identicalNo correction of the defect register changes this call.both outputsidentical: 0 of 10,813,440 values differ
replica, 1 thread − 8 threads (reproduce)Bit-identicalThe replica’s result does not depend on its thread count.both outputsidentical: 0 of 10,813,440 values differ

Δ is A − B over all 10,813,440 values of each output.

Provenance of this example
  • Data: OpenNeuro ds004958, sub-fsm64ps, CC0-1.0
  • Inputs: input 1 (sha256 84311fe0425f1bf6…)
  • Runs: figures cache entry ants-n4/panel-final-unit1, computed 2026-10-27T01:38:24+00:00 with larmorx 042d70d5a17c; ANTs v2.6.5 (fdce4d2f84) on ITK v5.4.5 (f51594ad88), x86_64-conda-linux-gnu-g++ (conda-forge gcc 14.3.0-20) 14.3.0; every command, its environment, threads and exit status are in its RUN_INFO.json

Every number above is in ants-n4.json, exported by python -m larmorx_validation evidence ants-n4.

Every case

Each case of the two records, the standard comparison and the data panel, with what it tests, the command line, its verdict and the reason. Search, filter by category or verdict (or select a segment of the chart above), sort by a column; #case=<name> links to one case.

122 cases

CaseCategoryWhat it testsVerdictChecks
fmriprep/boldref-coregistrationfMRIPrep's callsfMRIPrep's BOLD reference (niworkflows func/util.py:487): -d 3, b-spline distance 200, rescale (no effect without a mask), weight = premask (a blurred brain mask), ANTs' default shrink 4 (nipype drops shrink_factor=2); ds000005's boldref
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] -i openneuro-derivatives/ds000005-fmriprep/sub-01/func/sub-01_task-mixedgamblestask_run-1_boldref.nii.gz -o out.nii.gz -r --weight-image '<generated: n4-boldref-premask>'
Identical with correctly rounded maths

corrected: as built, 69747 of 139264 differ; max |diff| 0.00061, max rel 5.6e-06, 1 - NCC 5.4e-14 (÷ ANTs’ thread band: 0.43 max |Δ|, 0.45 1 − NCC); bit-identical with correctly rounded maths

12/12
fmriprep/boldref-templatefMRIPrep's callsthe same command on MNI152NLin2009cAsym's fMRIPrep boldref (res-02), weight = the brain probability map resampled onto it, as fMRIPrep's premask
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] -i templateflow/tpl-MNI152NLin2009cAsym/tpl-MNI152NLin2009cAsym_res-02_desc-fMRIPrep_boldref.nii.gz -o out.nii.gz -r --weight-image '<generated: n4-mni-premask>'
Identical with correctly rounded maths

corrected: as built, 467120 of 1082035 differ; max |diff| 0.00439, max rel 1.7e-06, 1 - NCC 0 (÷ ANTs’ thread band: 0.25 max |Δ|, 0 1 − NCC); bit-identical with correctly rounded maths

12/12
fmriprep/smriprep-pre-templatefMRIPrep's callssMRIPrep's N4 before the anatomical template (anatomical.py:1502): -d 3, ANTs' defaults (4 levels of 50 iterations, threshold 0, 4 control points, shrink 4); ds000005's raw T1w
command lineN4BiasFieldCorrection -d 3 --input-image openneuro/ds000005/sub-01/anat/sub-01_T1w.nii.gz --output out.nii.gz
Identical with correctly rounded maths

corrected: as built, 5413547 of 5898240 differ; max |diff| 0.241, max rel 0.00034, 1 - NCC 4.8e-10 (÷ ANTs’ thread band: 1.3 max |Δ|, 0.24 1 − NCC); bit-identical with correctly rounded maths

12/12
fmriprep/brain-extraction-pass1fMRIPrep's callssMRIPrep's brain extraction, first N4 (niworkflows anat/ants.py:243): -b [200] -c [50x50x50x50,1e-7] -s 4, on the truncated T1w
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50, 1e-07 ] --input-image '<generated: n4-t1-truncated>' --output out.nii.gz --shrink-factor 4
Bit-identical9/9
fmriprep/brain-extraction-final worked examplefMRIPrep's callssMRIPrep's final N4 (anat/ants.py:704): -b [200] -c [50x50x50x50x50,1e-7] -s 4 -r, weight = the WM probability map, bias field saved; ds000005's raw T1w
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --input-image openneuro/ds000005/sub-01/anat/sub-01_T1w.nii.gz --output [ out.nii.gz, bias.nii.gz ] -r --shrink-factor 4 --weight-image openneuro-derivatives/ds000005-fmriprep/sub-01/anat/sub-01_label-WM_probseg.nii.gz
Bit-identical14/14
fmriprep/n4-onlyfMRIPrep's callssMRIPrep's N4 for skull-stripped T1w images (anat/ants.py:927): the final N4's options without a weight image, on a brain-masked T1w
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --input-image '<generated: n4-t1-brain>' --output [ out.nii.gz, bias.nii.gz ] -r --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 694333 of 5898240 differ; max |diff| 0.0501, max rel 8.3e-05, 1 - NCC 1.3e-10 (÷ ANTs’ thread band: 0.25 max |Δ|, 0.01 1 − NCC); bias field: as built, 5898130 of 5898240 differ; max |diff| 7.86e-05, max rel 8.3e-05, 1 - NCC 2.1e-08 (÷ ANTs’ thread band: 0.15 max |Δ|, 0.0053 1 − NCC); bit-identical with correctly rounded maths

20/20
fmriprep/sdcflows-magnitudefMRIPrep's callssdcflows' brain extraction of a fieldmap magnitude (workflows/ancillary.py:87): -c [50x50x50x50x50,1e-7] -s 4; ds003012's magnitude1
command lineN4BiasFieldCorrection -d 3 --convergence [ 50x50x50x50x50, 1e-07 ] --input-image openneuro/ds003012/sub-01/fmap/sub-01_magnitude1.nii.gz --output out.nii.gz --shrink-factor 4
Bit-identical9/9
fmriprep/sdcflows-magnitude-ds001600fMRIPrep's callsthe same command on ds001600's magnitude1 (acq-v4)
command lineN4BiasFieldCorrection -d 3 --convergence [ 50x50x50x50x50, 1e-07 ] --input-image openneuro/ds001600/sub-1/fmap/sub-1_acq-v4_magnitude1.nii.gz --output out.nii.gz --shrink-factor 4
Bit-identical9/9
fmriprep/t1w-ds000210-finalfMRIPrep's callsthe final N4's options (without weights) on ds000210's T1w
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --input-image openneuro/ds000210/sub-02/anat/sub-02_T1w.nii.gz --output [ out.nii.gz, bias.nii.gz ] --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 6707886 of 11534336 differ; max |diff| 0.452, max rel 0.00017, 1 - NCC 6.9e-11 (÷ ANTs’ thread band: 0.21 max |Δ|, 0.064 1 − NCC); bias field: as built, 11508335 of 11534336 differ; max |diff| 0.000274, max rel 0.00017, 1 - NCC 4.8e-09 (÷ ANTs’ thread band: 0.21 max |Δ|, 0.17 1 − NCC); bit-identical with correctly rounded maths

20/20
fmriprep/boldref-maskfMRIPrep's callsds000005's boldref with its brain mask: outside-mask voxels restored, the rest rescaled to the input's range in the mask (ANTs' default with -x); bias field saved
command lineN4BiasFieldCorrection -d 3 -b [200] -i openneuro-derivatives/ds000005-fmriprep/sub-01/func/sub-01_task-mixedgamblestask_run-1_boldref.nii.gz -x openneuro-derivatives/ds000005-fmriprep/sub-01/func/sub-01_task-mixedgamblestask_run-1_desc-brain_mask.nii.gz -o [out.nii.gz,bias.nii.gz]
Bit-identical14/14
fmriprep/t1w-preproc-maskfMRIPrep's callsfMRIPrep's preprocessed T1w with -x the brain mask, -r 0, shrink 2
command lineN4BiasFieldCorrection -d 3 -i openneuro-derivatives/ds000005-fmriprep/sub-01/anat/sub-01_desc-preproc_T1w.nii.gz -x openneuro-derivatives/ds000005-fmriprep/sub-01/anat/sub-01_desc-brain_mask.nii.gz -r 0 -s 2 -c [20x20x20] -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 517729 of 5898240 differ; max |diff| 0.000244, max rel 3.4e-07, 1 - NCC 5.8e-14 (÷ ANTs’ thread band: 0.023 max |Δ|, 0.015 1 − NCC); bit-identical with correctly rounded maths

12/12
options/defaultsoptionsno option but input and output
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -o out.nii.gz
Bit-identical9/9
options/maskoptionsa binary mask: restoration outside, rescaling inside; bias field
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -x '<generated: brain-uint8>' -o [out.nii.gz,bias.nii.gz]
Bit-identical14/14
options/mask-no-rescaleoptionsa mask with -r 0
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -x '<generated: brain-uint8>' -r 0 -o out.nii.gz
Bit-identical9/9
options/fractional-maskoptionsa float mask (0, 0.3, 0.6, 0.999, 1, 1.7, 2): N4 uses every non-zero voxel; only voxels within 4 ulps of 1 are rescaled
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -x '<generated: fractional-float32>' -o out.nii.gz
Bit-identical9/9
options/weightoptionsa probability map as weight image (exact 0, 0.5 and 1)
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -w '<generated: gm-probseg>' -o [out.nii.gz,bias.nii.gz]
Bit-identical14/14
options/mask-and-weightoptionsmask and weight image together
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -x '<generated: brain-uint8>' -w '<generated: gm-probseg>' -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 10838 of 35840 differ; max |diff| 0.0612, max rel 0.00056, 1 - NCC 1.6e-08 (÷ ANTs’ thread band: 0.87 max |Δ|, 0.79 1 − NCC); bit-identical with correctly rounded maths

12/12
options/shrink-1options-s 1: no subsampling
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -s 1 -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 33414 of 35840 differ; max |diff| 0.000122, max rel 8e-07, 1 - NCC 4.4e-15 (÷ ANTs’ thread band: 0.36 max |Δ|, 0.26 1 − NCC); bit-identical with correctly rounded maths

12/12
options/shrink-3options-s 3
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -s 3 -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 35840 of 35840 differ; max |diff| 0.133, max rel 0.0063, 1 - NCC 3.5e-09 (÷ ANTs’ thread band: 1.7 max |Δ|, 4.2 1 − NCC); bit-identical with correctly rounded maths

12/12
options/shrink-0options-s 0: ITK makes it 1
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -s 0 -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 33414 of 35840 differ; max |diff| 0.000122, max rel 8e-07, 1 - NCC 4.4e-15 (÷ ANTs’ thread band: 0.36 max |Δ|, 0.26 1 − NCC); bit-identical with correctly rounded maths

12/12
options/distance-30options-b [30]: padding to whole spans of 30 mm, 2 or 3 spans per axis
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -b [30] -s 2 -o [out.nii.gz,bias.nii.gz]
Identical with correctly rounded maths

corrected: as built, 35824 of 35840 differ; max |diff| 0.0275, max rel 0.0039, 1 - NCC 3.8e-10 (÷ ANTs’ thread band: 0.4 max |Δ|, 0.058 1 − NCC); bias field: as built, 35827 of 35840 differ; max |diff| 0.00427, max rel 0.0039, 1 - NCC 2e-07 (÷ ANTs’ thread band: 0.4 max |Δ|, 0.092 1 − NCC); bit-identical with correctly rounded maths

20/20
options/distance-order-2options-b [25,2]: quadratic splines
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -b [25,2] -s 2 -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 35838 of 35840 differ; max |diff| 1.49, max rel 0.072, 1 - NCC 5.4e-08 (÷ ANTs’ thread band: 0.22 max |Δ|, 0.0028 1 − NCC); bit-identical with correctly rounded maths

12/12
options/meshoptions-b [2x3x1]: a mesh of 2 x 3 x 1 elements
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -b [2x3x1] -o out.nii.gz
Bit-identical9/9
options/order-5options-b [1x1x1,5]: quintic splines (Cox-de Boor kernels)
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -b [1x1x1,5] -c [10x10] -o out.nii.gz
Bit-identical9/9
options/convergence-stopoptions-c [30x30,0.001]: levels end as soon as the convergence value is not above 0.001
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -c [30x30,0.001] -o out.nii.gz
Bit-identical9/9
options/one-leveloptions-c [15]: one fitting level, no refinement
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -c [15] -o out.nii.gz
Bit-identical9/9
options/no-bracketsoptions-c 50x50 and -b 200 without brackets set nothing: ITK's one level of 50 iterations with threshold 0.001, 4 control points
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -c 50x50 -b 200 -o out.nii.gz
Bit-identical9/9
options/histogram-100options-t [0.3,0.02,100]: 100 bins (FFT of 256)
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -t [0.3,0.02,100] -c [20x20] -o out.nii.gz
Bit-identical9/9
options/histogram-256options-t [0.15,0.01,256]: 256 bins; log(256)/log(2) rounds above 8, so the FFT has 1024 points
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -t [0.15,0.01,256] -c [20x20] -o out.nii.gz
Bit-identical9/9
options/histogram-600options-t [0.15,0.01,600]: an FFT of 2048 points, where glibc's sincos is not correctly rounded for 4 twiddle factors
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -t [0.15,0.01,600] -c [20x20] -o out.nii.gz
Bit-identical9/9
options/output-one-parameteroptions-o [out]: one bracketed name
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -c [10] -o [out.nii.gz]
Bit-identical9/9
options/missing-maskoptionsa mask that does not exist: ones are used, but the mask counts as given (restoration and rescaling run)
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -x none.nii.gz -c [10x10] -o out.nii.gz
Bit-identical9/9
options/verboseoptions-v 1 (stdout is not compared: it holds timings and ITK's object dump)
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -v 1 -c [10x10] -o out.nii.gz
Bit-identical9/9
inputs/int16-scaledinputsscaled int16 input
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-int16-scaled>' -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 35836 of 35840 differ; max |diff| 0.11, max rel 0.00096, 1 - NCC 5.5e-09 (÷ ANTs’ thread band: 0.072 max |Δ|, 0.0031 1 − NCC); bit-identical with correctly rounded maths

12/12
inputs/zero-background-knowninputsa brain on an exact 0 background with 1 % dropout (0 and -20), no mask: by default the non-positive voxels are left out of the fit, which gives ANTs' own bias field with a mask of the positive voxels (one thread, correctly rounded maths); ANTs fits them like tissue
command lineN4BiasFieldCorrection -d 3 -i '<generated: known-n4-zero-background>' -s 2 -c [20x20,0] -o [out.nii.gz,bias.nii.gz]
Deliberate correction

ANTS-40, corrected by default (decided 2026-10-23): the replica’s default, fix, gives the methodically correct result bit for bit, checked against ANTs’ own N4 with a mask of the positive voxels, where the defect cannot arise (one thread, correctly rounded maths); reproduce gives ANTs’s output bit for bit (the record’s comparison).

21/21
inputs/outliersinputshot, negative and zero voxels (NaN and -inf read as 0): ANTs keeps non-positive values in its log image (ANTS-40, corrected by default)
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-outliers>' -c [20x20] -o out.nii.gz
Bit-identical9/9
inputs/constantinputsa constant image (the histogram has no width)
command lineN4BiasFieldCorrection -d 3 -i '<generated: constant>' -c [5] -o out.nii.gz
Bit-identical9/9
inputs/rampinputsvalues increasing in memory order: ITK's histogram minimum is never set
command lineN4BiasFieldCorrection -d 3 -i '<generated: n4-ramp>' -c [10x10] -o out.nii.gz
Bit-identical9/9
inputs/las-sform-onlyinputsan sform-only LAS image
command lineN4BiasFieldCorrection -d 3 -i '<generated: las-sform-only>' -c [20x20] -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 35840 of 35840 differ; max |diff| 0.0402, max rel 0.00024, 1 - NCC 5.7e-11 (÷ ANTs’ thread band: 0.46 max |Δ|, 0.31 1 − NCC); bit-identical with correctly rounded maths

12/12
inputs/negative-pixdiminputsa qform-only image with a negative pixdim
command lineN4BiasFieldCorrection -d 3 -i '<generated: negative-pixdim-qform>' -c [20x20] -o out.nii.gz
Bit-identical9/9
inputs/dim2inputs-d 2 on a 2D image
command lineN4BiasFieldCorrection -d 2 -i '<generated: slice-2d>' -o [out.nii.gz,bias.nii.gz]
Bit-identical14/14
inputs/dim2-distanceinputs-d 2 with -b [20] and -s 2
command lineN4BiasFieldCorrection -d 2 -i '<generated: slice-2d>' -b [20] -s 2 -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 1280 of 1280 differ; max |diff| 0.2, max rel 0.0014, 1 - NCC 1.6e-08 (÷ ANTs’ thread band: 0.33 max |Δ|, 0.56 1 − NCC); bit-identical with correctly rounded maths

12/12
inputs/dim4inputs-d 4: the series shrunk along space only, a spline along time too
command lineN4BiasFieldCorrection -d 4 -i '<generated: n4-series-4d>' -s 2 -c [10x10] -o [out.nii.gz,bias.nii.gz]
Bit-identical14/14
inputs/dim3-of-4dinputs-d 3 on a 4D file: its first volume, with an identity direction
command lineN4BiasFieldCorrection -d 3 -i '<generated: n4-series-4d>' -s 2 -c [10x10] -o out.nii.gz
Bit-identical9/9
inputs/dimension-from-fileinputsno -d: the dimension is the file's
command lineN4BiasFieldCorrection -i '<generated: head-float32>' -c [10] -o out.nii.gz
Bit-identical9/9
inputs/bold-first-volumeinputs-d 3 on fMRIPrep's preprocessed BOLD series: its first volume
command lineN4BiasFieldCorrection -d 3 -i openneuro-derivatives/ds000005-fmriprep/sub-01/func/sub-01_task-mixedgamblestask_run-1_space-T1w_desc-preproc_bold.nii.gz -b [200] -o out.nii.gz
Identical with correctly rounded maths

corrected: as built, 53416 of 94276 differ; max |diff| 0.00165, max rel 2.4e-06, 1 - NCC 2.4e-13 (÷ ANTs’ thread band: 0.39 max |Δ|, 0.73 1 − NCC); bit-identical with correctly rounded maths

12/12
errors/helperrors--help: the usage, exit 0
command lineN4BiasFieldCorrection --help
Bit-identical

exit status compared (both 0); the usage text is not compared

1/1
errors/shrink-not-integererrors-s 2.5: the parser's Convert<int> throws (ANTs aborts)
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -s 2.5 -o out.nii.gz
Rejected by both

ANTs exit -6: “s/antsCommandLineParser.h:133:\nITK ERROR: CommandLineParser(0x5bcba1094da0): ERROR: Parse error occured during command line argument processing\nERROR: Unable to convert ‘2.5’ to type ‘i’ as signed int”; the replica exit 1: "la…

1/1
errors/dimension-5errors-d 5: 'Unsupported dimension'
command lineN4BiasFieldCorrection -d 5 -i '<generated: head-float32>' -o out.nii.gz
Rejected by both

ANTs exit 1: ‘Unsupported dimension’; the replica exit 1: ‘Unsupported dimension’

1/1
errors/missing-inputerrorsan input that does not exist (ANTs crashes)
command lineN4BiasFieldCorrection -d 3 -i none.nii.gz -o out.nii.gz
Rejected by both

ANTs exit -11: ‘file /none.nii.gz does not exist .’; the replica exit 1: ‘file /none.nii.gz does not exist . \nlarmorx: N4BiasFieldCorrection: cannot read the input image (ANTs crashes here)’

1/1
errors/mesh-wrong-lengtherrors-b [2x2] in 3D: 'Incorrect mesh resolution', exit 1
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -b [2x2] -o out.nii.gz
Rejected by both

ANTs exit 1: ‘’; the replica exit 1: ‘’

1/1
errors/mask-other-griderrorsa mask on another grid: ITK throws (ANTs aborts)
command lineN4BiasFieldCorrection -d 3 -i '<generated: head-float32>' -x '<generated: n4-mask-other-grid>' -o out.nii.gz
Rejected by both

ANTs exit -6: ‘t occupy the same physical space! \nInputImage Origin: [2.1000000e+01, 3.0000000e+01, -2.0000000e+01], InputImagePrimary Origin: [2.4000000e+01, 3.0000000e+01, -2.0000000e+01]\n\tTolerance: 1.5000000e-06’; the replica exit 1: "l…

1/1
panel/fmriprep/ds000005-sub-01/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds000005-sub-01 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000005-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-01_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-01_T1w_noise_corrected_corrected.nii.gz, sub-01_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds000005-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 5327635 of 5898240 differ; max |diff| 0.0312, max rel 2.1e-05, 1 - NCC 3.3e-12 (÷ ANTs’ thread band: 1.6 max |Δ|, 2.1 1 − NCC); bias field: as built, 5842682 of 5898240 differ; max |diff| 5.78e-06, max rel 2.1e-05, 1 - NCC 4.4e-12 (÷ ANTs’ thread band: 1.2 max |Δ|, 1.4 1 − NCC); bit-identical with correctly rounded maths

20/20
panel/fmriprep/ds000005-sub-01/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds000005-sub-01 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000005-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-01_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-01_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 5815826 of 5898240 differ; max |diff| 0.0094, max rel 2.2e-05, 1 - NCC 3.3e-12 (÷ ANTs’ thread band: 0.85 max |Δ|, 0.33 1 − NCC); bit-identical with correctly rounded maths

12/12
panel/fmriprep/ds000005-sub-01/bold[task_mixedgamblestask_run_01]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_mixedgamblestask_run_01]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds000005-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000005-sub-01-noreconall/out/sub-01/func/sub-01_task-mixedgamblestask_run-01_desc-hmc_boldref.nii.gz' --output sub-01_task-mixedgamblestask_run-01_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds000005-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/bold_task_mixedgamblestask_run_01_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 76229 of 139264 differ; max |diff| 0.0114, max rel 1.8e-05, 1 - NCC 2.5e-12 (÷ ANTs’ thread band: 3.1 max |Δ|, 8.6 1 − NCC); bit-identical with correctly rounded maths

12/12
panel/fmriprep/ds000005-sub-01/bold[task_mixedgamblestask_run_02]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_mixedgamblestask_run_02]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds000005-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000005-sub-01-noreconall/out/sub-01/func/sub-01_task-mixedgamblestask_run-02_desc-hmc_boldref.nii.gz' --output sub-01_task-mixedgamblestask_run-02_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds000005-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/bold_task_mixedgamblestask_run_02_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 137886 of 139264 differ; max |diff| 0.381, max rel 0.00076, 1 - NCC 1.9e-09 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds000005-sub-01/bold[task_mixedgamblestask_run_03]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_mixedgamblestask_run_03]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds000005-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000005-sub-01-noreconall/out/sub-01/func/sub-01_task-mixedgamblestask_run-03_desc-hmc_boldref.nii.gz' --output sub-01_task-mixedgamblestask_run-03_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds000005-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/bold_task_mixedgamblestask_run_03_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 138068 of 139264 differ; max |diff| 1.27, max rel 0.0033, 1 - NCC 5.6e-08 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds000228-sub-pixar008/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds000228-sub-pixar008 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000228-sub-pixar008-noreconall/work/fmriprep_25_2_wf/sub_pixar008_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-pixar008_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-pixar008_T1w_noise_corrected_corrected.nii.gz, sub-pixar008_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds000228-sub-pixar008-noreconall/work/fmriprep_25_2_wf/sub_pixar008_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 4073942 of 6488064 differ; max |diff| 0.00439, max rel 8.5e-06, 1 - NCC 1.5e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 6460236 of 6488064 differ; max |diff| 8.46e-06, max rel 8.5e-06, 1 - NCC 6.6e-11 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds000228-sub-pixar008/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds000228-sub-pixar008 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000228-sub-pixar008-noreconall/work/fmriprep_25_2_wf/sub_pixar008_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-pixar008_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-pixar008_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 6364257 of 6488064 differ; max |diff| 0.00192, max rel 5.1e-06, 1 - NCC 5.7e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds000228-sub-pixar008/bold[task_pixar]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_pixar]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds000228-sub-pixar008 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000228-sub-pixar008-noreconall/out/sub-pixar008/func/sub-pixar008_task-pixar_desc-hmc_boldref.nii.gz' --output sub-pixar008_task-pixar_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds000228-sub-pixar008-noreconall/work/fmriprep_25_2_wf/sub_pixar008_wf/bold_task_pixar_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds000256-sub-CTS203/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds000256-sub-CTS203 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/work/fmriprep_25_2_wf/sub_CTS203_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-CTS203_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-CTS203_T1w_noise_corrected_corrected.nii.gz, sub-CTS203_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/work/fmriprep_25_2_wf/sub_CTS203_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 11205019 of 14680064 differ; max |diff| 0.000427, max rel 1.7e-06, 1 - NCC 4.2e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 14472799 of 14680064 differ; max |diff| 5.01e-06, max rel 1.6e-06, 1 - NCC 3.7e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds000256-sub-CTS203/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds000256-sub-CTS203 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/work/fmriprep_25_2_wf/sub_CTS203_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-CTS203_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-CTS203_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 13619162 of 14680064 differ; max |diff| 0.000153, max rel 5.1e-07, 1 - NCC 5.3e-15 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds000256-sub-CTS203/bold[task_restbaseline]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_restbaseline]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds000256-sub-CTS203 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/out/sub-CTS203/func/sub-CTS203_task-restbaseline_desc-hmc_boldref.nii.gz' --output sub-CTS203_task-restbaseline_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/work/fmriprep_25_2_wf/sub_CTS203_wf/bold_task_restbaseline_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds000256-sub-CTS203/bold[task_restfixationfixed]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_restfixationfixed]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds000256-sub-CTS203 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/out/sub-CTS203/func/sub-CTS203_task-restfixationfixed_desc-hmc_boldref.nii.gz' --output sub-CTS203_task-restfixationfixed_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/work/fmriprep_25_2_wf/sub_CTS203_wf/bold_task_restfixationfixed_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 127762 of 131072 differ; max |diff| 0.0161, max rel 5.4e-05, 1 - NCC 8.3e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds000256-sub-CTS203/bold[task_restfixationnofeedback]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_restfixationnofeedback]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds000256-sub-CTS203 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/out/sub-CTS203/func/sub-CTS203_task-restfixationnofeedback_desc-hmc_boldref.nii.gz' --output sub-CTS203_task-restfixationnofeedback_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds000256-sub-CTS203-noreconall/work/fmriprep_25_2_wf/sub_CTS203_wf/bold_task_restfixationnofeedback_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 127751 of 131072 differ; max |diff| 9.83, max rel 0.014, 1 - NCC 4.2e-07 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds001600-sub-1/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds001600-sub-1 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds001600-sub-1-noreconall/work/fmriprep_25_2_wf/sub_1_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-1_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-1_T1w_noise_corrected_corrected.nii.gz, sub-1_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds001600-sub-1-noreconall/work/fmriprep_25_2_wf/sub_1_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 4321765 of 11534336 differ; max |diff| 0.00244, max rel 2.1e-06, 1 - NCC 7.1e-14 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 9084014 of 11534336 differ; max |diff| 4.47e-07, max rel 2.1e-06, 1 - NCC 6.4e-14 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds001600-sub-1/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds001600-sub-1 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds001600-sub-1-noreconall/work/fmriprep_25_2_wf/sub_1_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-1_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-1_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 11534336 of 11534336 differ; max |diff| 0.00159, max rel 3.1e-06, 1 - NCC 1.9e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds001600-sub-1/fmap_preproc/wf_auto_00000/magnitude/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_auto_00000/magnitude/brainextraction/n4` of the fMRIPrep run ds001600-sub-1 (fmap_preproc_wf/wf_auto_00000/magnitude_wf/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds001600-sub-1-noreconall/work/fmriprep_25_2_wf/sub_1_wf/fmap_preproc_wf/wf_auto_00000/magnitude_wf/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 95243 of 180224 differ; max |diff| 0.00185, max rel 1.7e-05, 1 - NCC 2e-12 (÷ ANTs’ thread band: 0.7 max |Δ|, 0.45 1 − NCC); bit-identical with correctly rounded maths

12/12
panel/fmriprep/ds001600-sub-1/fmap_preproc/wf_auto_00001/magnitude/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_auto_00001/magnitude/brainextraction/n4` of the fMRIPrep run ds001600-sub-1 (fmap_preproc_wf/wf_auto_00001/magnitude_wf/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds001600-sub-1-noreconall/work/fmriprep_25_2_wf/sub_1_wf/fmap_preproc_wf/wf_auto_00001/magnitude_wf/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Bit-identical9/9
panel/fmriprep/ds001600-sub-1/fmap_preproc/wf_auto_00002/magnitude/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_auto_00002/magnitude/brainextraction/n4` of the fMRIPrep run ds001600-sub-1 (fmap_preproc_wf/wf_auto_00002/magnitude_wf/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds001600-sub-1-noreconall/work/fmriprep_25_2_wf/sub_1_wf/fmap_preproc_wf/wf_auto_00002/magnitude_wf/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Bit-identical9/9
panel/fmriprep/ds001600-sub-1/fmap_preproc/wf_auto_00003/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_auto_00003/brainextraction/n4` of the fMRIPrep run ds001600-sub-1 (fmap_preproc_wf/wf_auto_00003/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds001600-sub-1-noreconall/work/fmriprep_25_2_wf/sub_1_wf/fmap_preproc_wf/wf_auto_00003/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Bit-identical9/9
panel/fmriprep/ds002422-sub-02/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds002422-sub-02 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002422-sub-02-noreconall/work/fmriprep_25_2_wf/sub_02_wf/anat_fit_wf/anat_validate/sub-02_T1w_noise_corrected_ras_valid.nii' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-02_T1w_noise_corrected_ras_valid_corrected.nii, sub-02_T1w_noise_corrected_ras_valid_bias.nii ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds002422-sub-02-noreconall/work/fmriprep_25_2_wf/sub_02_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 8813882 of 11534336 differ; max |diff| 0.0278, max rel 1.2e-05, 1 - NCC 7e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 11533730 of 11534336 differ; max |diff| 4.35e-06, max rel 1.2e-05, 1 - NCC 4.5e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds002422-sub-02/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds002422-sub-02 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002422-sub-02-noreconall/work/fmriprep_25_2_wf/sub_02_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-02_T1w_noise_corrected_ras_valid_maths.nii' --convergence [ 50x50x50x50, 1e-07 ] --output sub-02_T1w_noise_corrected_ras_valid_maths_corrected.nii --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 11357777 of 11534336 differ; max |diff| 0.00537, max rel 6.4e-06, 1 - NCC 5.6e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds002422-sub-02/bold[task_arithm_run_01]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_arithm_run_01]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds002422-sub-02 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002422-sub-02-noreconall/out/sub-02/func/sub-02_task-arithm_run-01_desc-hmc_boldref.nii.gz' --output sub-02_task-arithm_run-01_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds002422-sub-02-noreconall/work/fmriprep_25_2_wf/sub_02_wf/bold_task_arithm_run_01_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 97325 of 98304 differ; max |diff| 3.6, max rel 0.0034, 1 - NCC 9.2e-07 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds002422-sub-02/bold[task_arithm_run_02]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_arithm_run_02]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds002422-sub-02 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002422-sub-02-noreconall/out/sub-02/func/sub-02_task-arithm_run-02_desc-hmc_boldref.nii.gz' --output sub-02_task-arithm_run-02_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds002422-sub-02-noreconall/work/fmriprep_25_2_wf/sub_02_wf/bold_task_arithm_run_02_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds002422-sub-02/bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds002422-sub-02 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002422-sub-02-noreconall/out/sub-02/func/sub-02_task-rest_desc-hmc_boldref.nii.gz' --output sub-02_task-rest_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds002422-sub-02-noreconall/work/fmriprep_25_2_wf/sub_02_wf/bold_task_rest_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds002422-sub-02/fmap_preproc/wf_auto_00000/magnitude/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_auto_00000/magnitude/brainextraction/n4` of the fMRIPrep run ds002422-sub-02 (fmap_preproc_wf/wf_auto_00000/magnitude_wf/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds002422-sub-02-noreconall/work/fmriprep_25_2_wf/sub_02_wf/fmap_preproc_wf/wf_auto_00000/magnitude_wf/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Bit-identical9/9
panel/fmriprep/ds002422-sub-02/fmap_preproc/wf_auto_00001/magnitude/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_auto_00001/magnitude/brainextraction/n4` of the fMRIPrep run ds002422-sub-02 (fmap_preproc_wf/wf_auto_00001/magnitude_wf/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds002422-sub-02-noreconall/work/fmriprep_25_2_wf/sub_02_wf/fmap_preproc_wf/wf_auto_00001/magnitude_wf/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 86612 of 147456 differ; max |diff| 0.00296, max rel 7.2e-06, 1 - NCC 3.2e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds002603-sub-01/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds002603-sub-01 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002603-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/anat_fit_wf/anat_validate/sub-01_acq-SaggitalNsig1Mm1_T1w_noise_corrected_ras_valid.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-01_acq-SaggitalNsig1Mm1_T1w_noise_corrected_ras_valid_corrected.nii.gz, sub-01_acq-SaggitalNsig1Mm1_T1w_noise_corrected_ras_valid_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds002603-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 4803389 of 16384000 differ; max |diff| 2.03, max rel 8.1e-06, 1 - NCC 8.4e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 16069774 of 16384000 differ; max |diff| 4.41e-06, max rel 8.1e-06, 1 - NCC 3.1e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds002603-sub-01/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds002603-sub-01 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002603-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-01_acq-SaggitalNsig1Mm1_T1w_noise_corrected_ras_valid_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-01_acq-SaggitalNsig1Mm1_T1w_noise_corrected_ras_valid_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 16383554 of 16384000 differ; max |diff| 38.9, max rel 0.00041, 1 - NCC 1.2e-09 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds002603-sub-01/bold[task_wm_acq_Mb4Mm3Tr550]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_wm_acq_Mb4Mm3Tr550]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds002603-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002603-sub-01-noreconall/out/sub-01/func/sub-01_task-wm_acq-Mb4Mm3Tr550_desc-hmc_boldref.nii.gz' --output sub-01_task-wm_acq-Mb4Mm3Tr550_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds002603-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/bold_task_wm_acq_Mb4Mm3Tr550_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 193271 of 230400 differ; max |diff| 0.141, max rel 3.6e-06, 1 - NCC 6.3e-14 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds002603-sub-01/bold[task_wm_acq_SeqMm3Tr2000]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_wm_acq_SeqMm3Tr2000]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds002603-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds002603-sub-01-noreconall/out/sub-01/func/sub-01_task-wm_acq-SeqMm3Tr2000_desc-hmc_boldref.nii.gz' --output sub-01_task-wm_acq-SeqMm3Tr2000_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds002603-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/bold_task_wm_acq_SeqMm3Tr2000_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds002603-sub-01/fmap_preproc/wf_auto_00000/magnitude/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_auto_00000/magnitude/brainextraction/n4` of the fMRIPrep run ds002603-sub-01 (fmap_preproc_wf/wf_auto_00000/magnitude_wf/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds002603-sub-01-noreconall/work/fmriprep_25_2_wf/sub_01_wf/fmap_preproc_wf/wf_auto_00000/magnitude_wf/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Bit-identical9/9
panel/fmriprep/ds003592-sub-280/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds003592-sub-280 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds003592-sub-280-noreconall/work/fmriprep_25_2_wf/sub_280_ses_1_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-280_ses-1_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-280_ses-1_T1w_noise_corrected_corrected.nii.gz, sub-280_ses-1_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds003592-sub-280-noreconall/work/fmriprep_25_2_wf/sub_280_ses_1_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 8779543 of 12582912 differ; max |diff| 0.000977, max rel 1.3e-06, 1 - NCC 2.1e-14 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 11970141 of 12582912 differ; max |diff| 7.75e-07, max rel 1.3e-06, 1 - NCC 4.7e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds003592-sub-280/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds003592-sub-280 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds003592-sub-280-noreconall/work/fmriprep_25_2_wf/sub_280_ses_1_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-280_ses-1_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-280_ses-1_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 12208723 of 12582912 differ; max |diff| 0.00146, max rel 2.7e-06, 1 - NCC 2.2e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds003592-sub-280/bold[ses_1_task_rest_echo_1]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[ses_1_task_rest_echo_1]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds003592-sub-280 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds003592-sub-280-noreconall/out/sub-280/ses-1/func/sub-280_ses-1_task-rest_desc-hmc_boldref.nii.gz' --output sub-280_ses-1_task-rest_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds003592-sub-280-noreconall/work/fmriprep_25_2_wf/sub_280_ses_1_wf/bold_ses_1_task_rest_echo_1_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds004173-sub-603309/anat_fit/anat_template/n4_correct[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/anat_template/n4_correct[0]` of the fMRIPrep run ds004173-sub-603309 (anat_fit_wf/anat_template_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds004173-sub-603309-noreconall/work/fmriprep_25_2_wf/sub_603309_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-603309_acq-headmotion1_T1w_noise_corrected.nii.gz' --output sub-603309_acq-headmotion1_T1w_noise_corrected_corrected.nii.gz
Identical with correctly rounded maths

corrected: as built, 9240964 of 12582912 differ; max |diff| 0.00269, max rel 3.4e-06, 1 - NCC 2e-13 (÷ ANTs’ thread band: 1.5 max |Δ|, 0.22 1 − NCC); bit-identical with correctly rounded maths

12/12
panel/fmriprep/ds004173-sub-603309/anat_fit/anat_template/n4_correct[1]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/anat_template/n4_correct[1]` of the fMRIPrep run ds004173-sub-603309 (anat_fit_wf/anat_template_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds004173-sub-603309-noreconall/work/fmriprep_25_2_wf/sub_603309_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise1/sub-603309_acq-headmotion2_T1w_noise_corrected.nii.gz' --output sub-603309_acq-headmotion2_T1w_noise_corrected_corrected.nii.gz
Identical with correctly rounded maths

corrected: as built, 4055291 of 12582912 differ; max |diff| 0.000183, max rel 5.2e-07, 1 - NCC 3.1e-15 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds004173-sub-603309/anat_fit/anat_template/n4_correct[2]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/anat_template/n4_correct[2]` of the fMRIPrep run ds004173-sub-603309 (anat_fit_wf/anat_template_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds004173-sub-603309-noreconall/work/fmriprep_25_2_wf/sub_603309_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise2/sub-603309_acq-standard_T1w_noise_corrected.nii.gz' --output sub-603309_acq-standard_T1w_noise_corrected_corrected.nii.gz
Identical with correctly rounded maths

corrected: as built, 5702824 of 12582912 differ; max |diff| 0.000244, max rel 5.7e-07, 1 - NCC 2.7e-15 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds004173-sub-603309/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds004173-sub-603309 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds004173-sub-603309-noreconall/work/fmriprep_25_2_wf/sub_603309_wf/anat_fit_wf/anat_template_wf/anat_merge/sub-603309_acq-headmotion1_T1w_noise_corrected_template.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-603309_acq-headmotion1_T1w_noise_corrected_template_corrected.nii.gz, sub-603309_acq-headmotion1_T1w_noise_corrected_template_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds004173-sub-603309-noreconall/work/fmriprep_25_2_wf/sub_603309_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 7252668 of 12582912 differ; max |diff| 0.00586, max rel 1.9e-06, 1 - NCC 1.2e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 7709761 of 12582912 differ; max |diff| 3.87e-07, max rel 1.9e-06, 1 - NCC 4.2e-14 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds004173-sub-603309/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds004173-sub-603309 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds004173-sub-603309-noreconall/work/fmriprep_25_2_wf/sub_603309_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-603309_acq-headmotion1_T1w_noise_corrected_template_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-603309_acq-headmotion1_T1w_noise_corrected_template_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 12582912 of 12582912 differ; max |diff| 0.00256, max rel 4.1e-06, 1 - NCC 1.4e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds004884-sub-M2079/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds004884-sub-M2079 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds004884-sub-M2079-noreconall/work/fmriprep_25_2_wf/sub_M2079_ses_217_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-M2079_ses-217_acq-tfl3p2_run-3_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-M2079_ses-217_acq-tfl3p2_run-3_T1w_noise_corrected_corrected.nii.gz, sub-M2079_ses-217_acq-tfl3p2_run-3_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds004884-sub-M2079-noreconall/work/fmriprep_25_2_wf/sub_M2079_ses_217_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 2583089 of 12582912 differ; max |diff| 0.0061, max rel 5.7e-06, 1 - NCC 9.2e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 12328289 of 12582912 differ; max |diff| 2.38e-06, max rel 5.7e-06, 1 - NCC 2.1e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds004884-sub-M2079/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds004884-sub-M2079 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds004884-sub-M2079-noreconall/work/fmriprep_25_2_wf/sub_M2079_ses_217_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-M2079_ses-217_acq-tfl3p2_run-3_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-M2079_ses-217_acq-tfl3p2_run-3_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 7337632 of 12582912 differ; max |diff| 0.000244, max rel 6.5e-07, 1 - NCC 1e-14 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds004884-sub-M2079/bold[ses_217_task_naming40_acq_epfid2_dir_AP_run_2]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[ses_217_task_naming40_acq_epfid2_dir_AP_run_2]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds004884-sub-M2079 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds004884-sub-M2079-noreconall/out/sub-M2079/ses-217/func/sub-M2079_ses-217_task-naming40_acq-epfid2_dir-AP_run-2_desc-hmc_boldref.nii.gz' --output sub-M2079_ses-217_task-naming40_acq-epfid2_dir-AP_run-2_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds004884-sub-M2079-noreconall/work/fmriprep_25_2_wf/sub_M2079_ses_217_wf/bold_ses_217_task_naming40_acq_epfid2_dir_AP_run_2_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds004884-sub-M2079/bold[ses_217_task_rest_acq_epfidp2p2_dir_AP_run_9]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[ses_217_task_rest_acq_epfidp2p2_dir_AP_run_9]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds004884-sub-M2079 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds004884-sub-M2079-noreconall/out/sub-M2079/ses-217/func/sub-M2079_ses-217_task-rest_acq-epfidp2p2_dir-AP_run-9_desc-hmc_boldref.nii.gz' --output sub-M2079_ses-217_task-rest_acq-epfidp2p2_dir-AP_run-9_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds004884-sub-M2079-noreconall/work/fmriprep_25_2_wf/sub_M2079_ses_217_wf/bold_ses_217_task_rest_acq_epfidp2p2_dir_AP_run_9_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds004958-sub-fsm64ps/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds004958-sub-fsm64ps (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds004958-sub-fsm64ps-noreconall/work/fmriprep_25_2_wf/sub_fsm64ps_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-fsm64ps_acq-mprageVnav_echo-1_part-mag_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-fsm64ps_acq-mprageVnav_echo-1_part-mag_T1w_noise_corrected_corrected.nii.gz, sub-fsm64ps_acq-mprageVnav_echo-1_part-mag_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds004958-sub-fsm64ps-noreconall/work/fmriprep_25_2_wf/sub_fsm64ps_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 11074293 of 11534336 differ; max |diff| 0.0439, max rel 1e-05, 1 - NCC 1.5e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 11534336 of 11534336 differ; max |diff| 4.95e-06, max rel 1e-05, 1 - NCC 8.2e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds004958-sub-fsm64ps/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds004958-sub-fsm64ps (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds004958-sub-fsm64ps-noreconall/work/fmriprep_25_2_wf/sub_fsm64ps_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-fsm64ps_acq-mprageVnav_echo-1_part-mag_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-fsm64ps_acq-mprageVnav_echo-1_part-mag_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 11486735 of 11534336 differ; max |diff| 0.0115, max rel 6.3e-06, 1 - NCC 2.9e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds004958-sub-fsm64ps/bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds004958-sub-fsm64ps (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image <workspace>/oracles/bids-views/ds004958/sub-fsm64ps/func/sub-fsm64ps_task-rest_sbref.nii.gz --output sub-fsm64ps_task-rest_sbref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds004958-sub-fsm64ps-noreconall/work/fmriprep_25_2_wf/sub_fsm64ps_wf/bold_task_rest_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds005003-sub-RS001/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds005003-sub-RS001 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005003-sub-RS001-noreconall/work/fmriprep_25_2_wf/sub_RS001_ses_1_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-RS001_ses-1_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-RS001_ses-1_T1w_noise_corrected_corrected.nii.gz, sub-RS001_ses-1_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds005003-sub-RS001-noreconall/work/fmriprep_25_2_wf/sub_RS001_ses_1_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 7487867 of 9437184 differ; max |diff| 0.00684, max rel 3.8e-06, 1 - NCC 9.7e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 9128052 of 9437184 differ; max |diff| 1.4e-06, max rel 3.8e-06, 1 - NCC 4.8e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds005003-sub-RS001/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds005003-sub-RS001 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005003-sub-RS001-noreconall/work/fmriprep_25_2_wf/sub_RS001_ses_1_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-RS001_ses-1_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-RS001_ses-1_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 9051972 of 9437184 differ; max |diff| 0.00409, max rel 1.4e-05, 1 - NCC 4.2e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005003-sub-RS001/bold[ses_1_task_motor]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[ses_1_task_motor]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds005003-sub-RS001 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005003-sub-RS001-noreconall/out/sub-RS001/ses-1/func/sub-RS001_ses-1_task-motor_desc-hmc_boldref.nii.gz' --output sub-RS001_ses-1_task-motor_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds005003-sub-RS001-noreconall/work/fmriprep_25_2_wf/sub_RS001_ses_1_wf/bold_ses_1_task_motor_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds005003-sub-RS001/bold[ses_1_task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[ses_1_task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds005003-sub-RS001 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005003-sub-RS001-noreconall/out/sub-RS001/ses-1/func/sub-RS001_ses-1_task-rest_desc-hmc_boldref.nii.gz' --output sub-RS001_ses-1_task-rest_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds005003-sub-RS001-noreconall/work/fmriprep_25_2_wf/sub_RS001_ses_1_wf/bold_ses_1_task_rest_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 145497 of 147456 differ; max |diff| 0.352, max rel 0.00062, 1 - NCC 1.3e-09 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005040-sub-002/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds005040-sub-002 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005040-sub-002-noreconall/work/fmriprep_25_2_wf/sub_002_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-002_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-002_T1w_noise_corrected_corrected.nii.gz, sub-002_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds005040-sub-002-noreconall/work/fmriprep_25_2_wf/sub_002_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 6301102 of 15842304 differ; max |diff| 0.422, max rel 2.6e-06, 1 - NCC 7.3e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 15735809 of 15842304 differ; max |diff| 1.25e-06, max rel 2.7e-06, 1 - NCC 1.2e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds005040-sub-002/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds005040-sub-002 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005040-sub-002-noreconall/work/fmriprep_25_2_wf/sub_002_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-002_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-002_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 15842295 of 15842304 differ; max |diff| 0.336, max rel 5.7e-06, 1 - NCC 2.3e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005040-sub-002/bold[task_clip_run_1]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_clip_run_1]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds005040-sub-002 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005040-sub-002-noreconall/out/sub-002/func/sub-002_task-clip_run-1_desc-hmc_boldref.nii.gz' --output sub-002_task-clip_run-1_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds005040-sub-002-noreconall/work/fmriprep_25_2_wf/sub_002_wf/bold_task_clip_run_1_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Bit-identical9/9
panel/fmriprep/ds005040-sub-002/fmap_preproc/wf_auto_00000/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_auto_00000/brainextraction/n4` of the fMRIPrep run ds005040-sub-002 (fmap_preproc_wf/wf_auto_00000/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds005040-sub-002-noreconall/work/fmriprep_25_2_wf/sub_002_wf/fmap_preproc_wf/wf_auto_00000/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Bit-identical9/9
panel/fmriprep/ds005454-sub-16/anat_fit/anat_template/n4_correct[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/anat_template/n4_correct[0]` of the fMRIPrep run ds005454-sub-16 (anat_fit_wf/anat_template_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds005454-sub-16-noreconall/work/fmriprep_25_2_wf/sub_16_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-16_acq-denoised_T1w_noise_corrected.nii.gz' --output sub-16_acq-denoised_T1w_noise_corrected_corrected.nii.gz
Identical with correctly rounded maths

corrected: as built, 10778341 of 17547264 differ; max |diff| 0.00708, max rel 3.2e-06, 1 - NCC 3e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005454-sub-16/anat_fit/anat_template/n4_correct[1]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/anat_template/n4_correct[1]` of the fMRIPrep run ds005454-sub-16 (anat_fit_wf/anat_template_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds005454-sub-16-noreconall/work/fmriprep_25_2_wf/sub_16_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise1/sub-16_acq-mp2rage_T1w_noise_corrected.nii.gz' --output sub-16_acq-mp2rage_T1w_noise_corrected_corrected.nii.gz
Identical with correctly rounded maths

corrected: as built, 12380266 of 17547264 differ; max |diff| 0.0547, max rel 8.6e-06, 1 - NCC 1.6e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005454-sub-16/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds005454-sub-16 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005454-sub-16-noreconall/work/fmriprep_25_2_wf/sub_16_wf/anat_fit_wf/anat_template_wf/anat_merge/sub-16_acq-denoised_T1w_noise_corrected_template.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-16_acq-denoised_T1w_noise_corrected_template_corrected.nii.gz, sub-16_acq-denoised_T1w_noise_corrected_template_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds005454-sub-16-noreconall/work/fmriprep_25_2_wf/sub_16_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 12470213 of 17547264 differ; max |diff| 0.033, max rel 1.5e-05, 1 - NCC 5.6e-12 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 12476998 of 17547264 differ; max |diff| 1.45e-05, max rel 1.5e-05, 1 - NCC 0.0027 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds005454-sub-16/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds005454-sub-16 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005454-sub-16-noreconall/work/fmriprep_25_2_wf/sub_16_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-16_acq-denoised_T1w_noise_corrected_template_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-16_acq-denoised_T1w_noise_corrected_template_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 17540037 of 17547264 differ; max |diff| 0.00439, max rel 2.7e-06, 1 - NCC 2.9e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005454-sub-16/bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds005454-sub-16 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005454-sub-16-noreconall/out/sub-16/func/sub-16_task-rest_desc-hmc_boldref.nii.gz' --output sub-16_task-rest_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds005454-sub-16-noreconall/work/fmriprep_25_2_wf/sub_16_wf/bold_task_rest_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 1982438 of 2457600 differ; max |diff| 0.000488, max rel 7.1e-07, 1 - NCC 5e-15 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005454-sub-16/fmap_preproc/wf_gre16/magnitude/brainextraction/n4data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `fmap_preproc/wf_gre16/magnitude/brainextraction/n4` of the fMRIPrep run ds005454-sub-16 (fmap_preproc_wf/wf_gre16/magnitude_wf/brainextraction_wf/n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection -d 3 --input-image '<fMRIPrep run>/ds005454-sub-16-noreconall/work/fmriprep_25_2_wf/sub_16_wf/fmap_preproc_wf/wf_gre16/magnitude_wf/brainextraction_wf/clipper_pre/clipped.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output clipped_corrected.nii.gz --shrink-factor 4
Bit-identical9/9
panel/fmriprep/ds005752-sub-ON01016/anat_fit/brain_extraction/atropos/inu_n4_final[0]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/atropos/inu_n4_final[0]` of the fMRIPrep run ds005752-sub-ON01016 (anat_fit_wf/brain_extraction_wf/atropos_wf/inu_n4_final): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005752-sub-ON01016-noreconall/work/fmriprep_25_2_wf/sub_ON01016_ses_01_wf/anat_fit_wf/anat_template_wf/denoise/mapflow/_denoise0/sub-ON01016_ses-01_acq-FSPGR_rec-SCIC_T1w_noise_corrected.nii.gz' --convergence [ 50x50x50x50x50, 1e-07 ] --output [ sub-ON01016_ses-01_acq-FSPGR_rec-SCIC_T1w_noise_corrected_corrected.nii.gz, sub-ON01016_ses-01_acq-FSPGR_rec-SCIC_T1w_noise_corrected_bias.nii.gz ] -r --shrink-factor 4 --weight-image '<fMRIPrep run>/ds005752-sub-ON01016-noreconall/work/fmriprep_25_2_wf/sub_ON01016_ses_01_wf/anat_fit_wf/brain_extraction_wf/atropos_wf/apply_wm_prior/prodmap.nii.gz'
Identical with correctly rounded maths

corrected: as built, 4316144 of 12845056 differ; max |diff| 0.0957, max rel 3.5e-05, 1 - NCC 6.6e-11 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bias field: as built, 11033687 of 12845056 differ; max |diff| 1.35e-05, max rel 3.5e-05, 1 - NCC 5.8e-11 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

16/16
panel/fmriprep/ds005752-sub-ON01016/anat_fit/brain_extraction/inu_n4[40]data panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `anat_fit/brain_extraction/inu_n4[40]` of the fMRIPrep run ds005752-sub-ON01016 (anat_fit_wf/brain_extraction_wf/inu_n4): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005752-sub-ON01016-noreconall/work/fmriprep_25_2_wf/sub_ON01016_ses_01_wf/anat_fit_wf/brain_extraction_wf/truncate_images/mapflow/_truncate_images0/sub-ON01016_ses-01_acq-FSPGR_rec-SCIC_T1w_noise_corrected_maths.nii.gz' --convergence [ 50x50x50x50, 1e-07 ] --output sub-ON01016_ses-01_acq-FSPGR_rec-SCIC_T1w_noise_corrected_maths_corrected.nii.gz --shrink-factor 4
Identical with correctly rounded maths

corrected: as built, 12845056 of 12845056 differ; max |diff| 0.00989, max rel 6.1e-06, 1 - NCC 6.3e-13 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005752-sub-ON01016/bold[ses_01_task_rest_dir_forward]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[ses_01_task_rest_dir_forward]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds005752-sub-ON01016 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005752-sub-ON01016-noreconall/out/sub-ON01016/ses-01/func/sub-ON01016_ses-01_task-rest_dir-forward_desc-hmc_boldref.nii.gz' --output sub-ON01016_ses-01_task-rest_dir-forward_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds005752-sub-ON01016-noreconall/work/fmriprep_25_2_wf/sub_ON01016_ses_01_wf/bold_ses_01_task_rest_dir_forward_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 193141 of 196608 differ; max |diff| 3.3, max rel 0.0018, 1 - NCC 5.4e-09 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/fmriprep/ds005752-sub-ON01016/bold[ses_01_task_rest_dir_reverse]/bold_fit/enhance_and_skullstrip_bold/n4_correctdata panel: fMRIPrep's callsfMRIPrep 25.2.6's N4BiasFieldCorrection node `bold[ses_01_task_rest_dir_reverse]/bold_fit/enhance_and_skullstrip_bold/n4_correct` of the fMRIPrep run ds005752-sub-ON01016 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/n4_correct): its command.txt replayed, its inputs from the run's working directory
command lineN4BiasFieldCorrection --bspline-fitting [ 200 ] -d 3 --input-image '<fMRIPrep run>/ds005752-sub-ON01016-noreconall/out/sub-ON01016/ses-01/func/sub-ON01016_ses-01_task-rest_dir-reverse_desc-hmc_boldref.nii.gz' --output sub-ON01016_ses-01_task-rest_dir-reverse_desc-hmc_boldref_corrected.nii.gz -r --weight-image '<fMRIPrep run>/ds005752-sub-ON01016-noreconall/work/fmriprep_25_2_wf/sub_ON01016_ses_01_wf/bold_ses_01_task_rest_dir_reverse_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/fix_header/tpl-MNI152NLin2009cAsym_res-01_label-brain_probseg_trans_fixhdr.nii.gz'
Identical with correctly rounded maths

corrected: as built, 192977 of 196608 differ; max |diff| 2.84, max rel 0.00079, 1 - NCC 3.5e-09 (÷ ANTs’ thread band: – max |Δ|, – 1 − NCC); bit-identical with correctly rounded maths

10/10
panel/images/final/7t-mp2ragedata panel: other imagessMRIPrep's final N4 options (-b [200] -c [50x50x50x50x50,1e-7] -s 4, the bias field saved) on the raw 7 T MP2RAGE: the panel's largest anatomical image, a noisy background as bright as the head (ds005454 sub-16: 7 T MP2RAGE uniform image as reconstructed (0.8 mm, 224 x 272 x 288, its noisy background kept), the panel's largest anatomical image)
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image openneuro/ds005454/sub-16/anat/sub-16_acq-mp2rage_T1w.nii.gz --output [ out.nii.gz, bias.nii.gz ]
Identical with correctly rounded maths

corrected: as built, 12354443 of 17547264 differ; max |diff| 0.0413, max rel 1.1e-05, 1 - NCC 2e-12 (÷ ANTs’ thread band: 0.34 max |Δ|, 0.26 1 − NCC); bias field: as built, 17503900 of 17547264 differ; max |diff| 9.42e-06, max rel 1.1e-05, 1 - NCC 1.1e-10 (÷ ANTs’ thread band: 0.39 max |Δ|, 0.23 1 − NCC); bit-identical with correctly rounded maths

20/20
panel/images/final/3t-unit1data panel: other imagesthe same options on a 3 T MP2RAGE UNIT1 (its salt-and-pepper background) (ds004958 sub-fsm64ps: 3 T MP2RAGE UNIT1 of a 68-year-old (1 mm, its salt-and-pepper background))
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image openneuro/ds004958/sub-fsm64ps/anat/sub-fsm64ps_acq-mp2rage_UNIT1.nii.gz --output [ out.nii.gz, bias.nii.gz ]
Identical with correctly rounded maths

corrected: as built, 10713781 of 10813440 differ; max |diff| 0.0151, max rel 3.2e-06, 1 - NCC 2.1e-12 (÷ ANTs’ thread band: 0.14 max |Δ|, 0.58 1 − NCC); bias field: as built, 10803732 of 10813440 differ; max |diff| 4.05e-06, max rel 3.2e-06, 1 - NCC 4.8e-12 (÷ ANTs’ thread band: 0.2 max |Δ|, 0.54 1 − NCC); bit-identical with correctly rounded maths

20/20
panel/images/final/flairdata panel: other imagesthe same options on a FLAIR of 3 mm slices (shrink 4: 11 slices) (ds003592 sub-280: an 82-year-old's FLAIR (0.42 x 0.42 x 3 mm, 44 slices))
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image openneuro/ds003592/sub-280/ses-1/anat/sub-280_ses-1_FLAIR.nii.gz --output [ out.nii.gz, bias.nii.gz ]
Bit-identical14/14
panel/images/final/tumour-t2wdata panel: other imagesthe same options on a T2w of 6 mm slices (shrink 4: 7 slices) with a tumour (ds005003 sub-RS001: the tumour's T2w: 5 mm slices 6 mm apart (0.72 x 0.72 x 6 mm))
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image openneuro/ds005003/sub-RS001/ses-1/anat/sub-RS001_ses-1_T2w.nii.gz --output [ out.nii.gz, bias.nii.gz ]
Identical with correctly rounded maths

corrected: as built, 2374386 of 3072000 differ; max |diff| 0.0192, max rel 3.8e-05, 1 - NCC 6.2e-12 (÷ ANTs’ thread band: 0.058 max |Δ|, 0.048 1 − NCC); bias field: as built, 3071975 of 3072000 differ; max |diff| 7.15e-05, max rel 3.9e-05, 1 - NCC 6e-10 (÷ ANTs’ thread band: 0.12 max |Δ|, 0.024 1 − NCC); bit-identical with correctly rounded maths

20/20
panel/images/final-mask/stroke-t2wdata panel: other imagesthe same options on a stroke's T2w, -x the head without the manual lesion mask (N4 fitted outside the lesion) (ds004884 sub-M2079: a chronic left-hemisphere stroke (176 mL) on its T2w; ds004884 sub-M2079: the stroke's manual lesion mask, drawn on the T2w (uint8))
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image openneuro/ds004884/sub-M2079/ses-217/anat/sub-M2079_ses-217_acq-spc3_run-4_T2w.nii.gz --output [ out.nii.gz, bias.nii.gz ] --mask-image '<generated: panel-stroke-nonlesion-mask>'
Bit-identical14/14
panel/images/final-mask/stroke-t2w-mask-from-affinedata panel: other imagesthe same with the mask written from the T2w's affine (its spacing recomputed, 2e-6 mm off): ITK's VerifyInputInformation throws, ANTs aborts and the replica refuses (ds004884 sub-M2079: a chronic left-hemisphere stroke (176 mL) on its T2w; ds004884 sub-M2079: the stroke's manual lesion mask, drawn on the T2w (uint8))
command lineN4BiasFieldCorrection -d 3 --bspline-fitting [ 200 ] --convergence [ 50x50x50x50x50, 1e-07 ] --shrink-factor 4 --input-image openneuro/ds004884/sub-M2079/ses-217/anat/sub-M2079_ses-217_acq-spc3_run-4_T2w.nii.gz --output [ out.nii.gz, bias.nii.gz ] --mask-image '<generated: panel-stroke-nonlesion-mask-affine>'
Rejected by both

ANTs exit -6: ‘t occupy the same physical space! \nInputImage Spacing: [9.9999797e-01, 1.0000000e+00, 1.0000000e+00], InputImagePrimary Spacing: [1.0000000e+00, 1.0000000e+00, 1.0000000e+00]\n\tTolerance: 9.9999797e-07’; the replica exit 1: "l…

1/1

Speed

The wall time of the whole command on six command lines of fMRIPrep, sMRIPrep and sdcflows, from a BOLD reference to a raw T1w, with ANTs and the replica at 1 and 12 threads (benchmark).

Speed, the replica against the original: ANTs N4BiasFieldCorrection 2.6.5

Seconds for the whole command (reading, N4, writing), the median of repeated runs, on a log scale. The factor is how many times faster the replica is than ANTs, both on one thread.

  • ANTs N4BiasFieldCorrection, 1 thread
  • ANTs N4BiasFieldCorrection, 12 threads
  • replica, 1 thread
  • replica, 12 threads
  1. BOLD reference, ds000005ANTs N4BiasFieldCorrection, 1 thread 1 sANTs N4BiasFieldCorrection, 12 threads 0.556 sreplica, 1 thread 0.135 sreplica, 12 threads 0.0993 s7.4×
  2. BOLD reference, MNI templateANTs N4BiasFieldCorrection, 1 thread 3.89 sANTs N4BiasFieldCorrection, 12 threads 1.24 sreplica, 1 thread 0.534 sreplica, 12 threads 0.278 s7.3×
  3. T1w, before the templateANTs N4BiasFieldCorrection, 1 thread 55.2 sANTs N4BiasFieldCorrection, 12 threads 10.7 sreplica, 1 thread 8.69 sreplica, 12 threads 2.4 s6.4×
  4. T1w, brain extraction, first N4ANTs N4BiasFieldCorrection, 1 thread 57.9 sANTs N4BiasFieldCorrection, 12 threads 11.3 sreplica, 1 thread 8.59 sreplica, 12 threads 2.69 s6.7×
  5. T1w, brain extraction, final N4ANTs N4BiasFieldCorrection, 1 thread 12.4 sANTs N4BiasFieldCorrection, 12 threads 4.33 sreplica, 1 thread 1.79 sreplica, 12 threads 0.69 s6.9×
  6. field-map magnitude, ds003012ANTs N4BiasFieldCorrection, 1 thread 1.96 sANTs N4BiasFieldCorrection, 12 threads 0.907 sreplica, 1 thread 0.271 sreplica, 12 threads 0.16 s7.2×
0.01 s0.1 s1 s10 s100 s

From the benchmark report docs/benchmarks/ants-n4.json (Intel® Core™ i7-8750H CPU @ 2.20GHz, 12 logical CPUs), measured on 2026-10-11 against the ANTs build of that day; a re-run against the repository’s current build is planned.

Corrections by default

Where ANTs has a deterministic defect that changes a result, the replica corrects it by default, fix, and warns. upstream_defects="reproduce" (on the command line --upstream-defects reproduce from the next release, LARMORX_UPSTREAM_DEFECTS=reproduce in 0.1.0rc1) gives ANTs’ behaviour, and the records compare with it. The record’s list of what fix changed meets ANTS-35, ANTS-36, ANTS-38, ANTS-40 and ITK-02 (register, findings). Of them, ANTS-40 is the one measured on fMRIPrep’s calls in the five runs (above); the measurement and its code are in the repository (investigation). Many of the data panel’s calls pass -r without a mask (ANTS-36): the replica, like ANTs, does nothing with it in either mode, and warns. The panel’s record names it in 36 cases. In the tumour example the values change through ANTS-40 alone.

Random testing

Random testing runs random valid inputs and option combinations through the original and the replica, in long runs of several hours: inputs from seeded plain arithmetic (up to 32³ voxels; orientations, data types, scale factors and value ranges of every kind), with options drawn in random combinations from those the program accepts (how random testing works). A difference is shrunk to the smallest input and fewest options that still show it, then fixed and kept as a case of the record, registered as a defect of the original, or reported. The latest long run:

Random testing, the replica against the original: ANTs N4BiasFieldCorrection 2.6.5, the long run of 2026-10-25 (seed 20261025, 6.2 h for 21 programs): random valid inputs and option combinations, each run through both

ProgramRandom casesExtremeOptions coveredAgreeglibcNoticesUndefinedDifferHeavyGuardUnclear
N4BiasFieldCorrectionreplica1,017011649·315··39·14

No difference.

Notices: a correction named in a case where it changed nothing the input shows (defect, cases): N4BiasFieldCorrection: ANTS-35 (180), ANTS-36 (50), ANTS-38 (19), ANTS-39 (1), ANTS-40 (102).

Heavy cases (a run over the memory limit of a case, queued and run to its end later, one at a time, in a large box): ants-n4:20261025:24 (beyond any machine: predicted, not run); ants-n4:20261025:46 (run to its end: agree, the run's peak 29.1 GiB); ants-n4:20261025:56 (beyond any machine: predicted, not run); ants-n4:20261025:98 (run to its end: agree, the run's peak 33.8 GiB); ants-n4:20261025:108 (beyond any machine: predicted, not run); ants-n4:20261025:151 (run to its end: notice, the run's peak 7.4 GiB); ants-n4:20261025:156 (beyond any machine: predicted, not run); ants-n4:20261025:161 (pending: larger machine: not run here); ants-n4:20261025:209 (run to its end: notice, the run's peak 12.9 GiB); ants-n4:20261025:245 (pending: larger machine: not run here); ants-n4:20261025:273 (run to its end: agree, the run's peak 5.4 GiB); ants-n4:20261025:275 (run to its end: agree, the run's peak 9.1 GiB); ants-n4:20261025:242 (run to its end: agree, the run's peak 4.7 GiB); ants-n4:20261025:285 (pending: larger machine: not run here); ants-n4:20261025:388 (pending: larger machine: not run here); ants-n4:20261025:404 (run to its end: agree, the run's peak 23.9 GiB); ants-n4:20261025:406 (run to its end: notice, the run's peak 0.9 GiB); ants-n4:20261025:458 (pending: larger machine: not run here); ants-n4:20261025:473 (run to its end: notice, the run's peak 3.2 GiB); ants-n4:20261025:475 (run to its end: notice, the run's peak 2.7 GiB); ants-n4:20261025:500 (run to its end: agree, the run's peak 3.4 GiB); ants-n4:20261025:510 (run to its end: notice, the run's peak 5.9 GiB); ants-n4:20261025:557 (pending: larger machine: not run here); ants-n4:20261025:560 (run to its end: notice, the run's peak 5.4 GiB); ants-n4:20261025:575 (run to its end: agree, the run's peak 18.5 GiB); ants-n4:20261025:630 (run to its end: agree, the run's peak 3.9 GiB); ants-n4:20261025:650 (run to its end: agree, the run's peak 0.0 GiB); ants-n4:20261025:621 (run to its end: agree, the run's peak 23.4 GiB); ants-n4:20261025:715 (run to its end: agree, the run's peak 9.0 GiB); ants-n4:20261025:724 (beyond any machine: predicted, not run); ants-n4:20261025:721 (run to its end: notice, the run's peak 5.2 GiB); ants-n4:20261025:751 (run to its end: agree, the run's peak 2.9 GiB); ants-n4:20261025:728 (pending: larger machine: not run here); ants-n4:20261025:750 (run to its end: agree, the run's peak 17.1 GiB); ants-n4:20261025:777 (run to its end: agree, the run's peak 46.0 GiB); ants-n4:20261025:849 (run to its end: agree, the run's peak 6.0 GiB); ants-n4:20261025:894 (run to its end: notice, the run's peak 5.0 GiB); ants-n4:20261025:899 (beyond any machine: predicted, not run); ants-n4:20261025:978 (run to its end: agree, the run's peak 21.4 GiB).

Every case is drawn from its tool, the run's seed and its number, and replays anywhere: how random testing works; each difference's triage: validation/difftest/triage.toml.

Known limits

The comparison with ANTs is made on Linux x86-64; the cross-platform checks (how they work) hold the other five platforms to LARMORX’s own Linux output. The ANTs it is compared with is built by GCC 14.3, which honours the source’s float roundings; an older GCC 11 build of ANTs gives other bits (findings).

History

  • 2026-10-12: ANTS-35, ANTS-36, ANTS-38 and ITK-02 corrected by default (register).
  • 2026-10-23: ANTS-40 measured on fMRIPrep’s 25 N4 calls, as the project’s rules for a defect of unknown effect require, and corrected by default.
  • 2026-10-26: the data panel’s record, every N4 call of the fourteen runs and five more images.

Data credits

The data this page shows or names: each dataset with its version, DOI and licence, from LARMORX's test-data catalog. Public-domain dedications (CC0, PDDL) ask nothing; the datasets are cited all the same.