Evidence · ANTs

antsAI and antsAffineInitializer

antsAI and antsAffineInitializer find the starting transform of a registration: antsAI from centres, principal axes or matched blobs and a search from many starting points, antsAffineInitializer from a search of rotations about the principal axes. LARMORX’s port of each is a replica of ANTs 2.6.5’s program on ITK 5.4.5. In reproduce they give ANTs’ exit status, messages and transform file byte for byte in every compared case of their records once ANTs uses correctly rounded maths: antsAI 76 of 76 and 40 of 40 on the data panel, antsAffineInitializer 26 of 26 and 14 of 14. With ANTs as built, 75 of antsAI’s 76 and 38 of its 40 match, and all of antsAffineInitializer’s; the others differ only where glibc’s maths is not correctly rounded. On fMRIPrep’s own brain extraction the replica writes ANTs’ one-thread transform. Their default, fix, corrects six defects of the blob alignment, ANTS-116 among them.

Ports
  • replica larmorx Apache-2.0: each program translated from ANTs 2.6.5's and ITK 5.4.5's source (Apache-2.0)
Status
replica validated, each program, with its data panel (marked before the statuses' definition)
Records
antsAI: standard comparison, full comparison, data panel; antsAffineInitializer: standard comparison, full comparison, data panel; all against ANTs 2.6.5 · benchmark ants-ai
Reference
antsAI, antsAffineInitializer · API pages ants-ai · ants-affine-initializer · findings ants-ai

What is compared

Both find a starting transform for a registration: antsAI aligns centres, principal axes or moments, or matched blobs (-b), and searches rotations and translations from many starting points against a metric (Mattes, GC, MI); antsAffineInitializer searches rotations about the principal axes. fMRIPrep’s init_aff steps run antsAI before the brain extraction and the BOLD-to-T1w registration, without blobs.

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

bash
larmorx ants antsAI -d 3 -m 'Mattes[template.nii.gz,t1w.nii.gz,32,Regular,0.25]' -x template_mask.nii.gz -t 'Affine[0.1]' -s '[15,0.1]' -g '[40.0,0x40x40]' -p 0 -c '[10,1e-06,10]' -o initialization.mat
larmorx ants antsAffineInitializer 3 fixed.nii.gz moving.nii.gz init.mat 15 0.1 0 10

In Python: lx.ants.ai (blobs=30 for -b 30), lx.ants.blob_alignment (the blob alignment’s every step) and lx.ants.affine_initializer. How to read what follows: how to read the evidence.

Their port

ANTs and ITK come under the Apache-2.0 licence, as LARMORX does, so the replicas are in larmorx itself. Each 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).

The replicas compute the C library’s maths correctly rounded, the same on every platform. Where glibc’s maths is not correctly rounded for a value a run meets, ANTs as built can write a transform that differs from theirs in its last bits.

The verdicts at a glance

Agreement, replica against the original: ANTs antsAI 2.6.5 (standard comparison and data panel, 172 cases of the two records), by category and verdict
initial alignments9
the multi-start search38
messages51
random sampling2
real images6
the blob alignment25
random testing1
data panel: fMRIPrep's calls36
data panel: other images4

Select a segment to list its cases below.

Agreement, replica against the original: ANTs antsAffineInitializer 2.6.5 (standard comparison and data panel, 55 cases of the two records), by category and verdict
messages19
the multi-start search21
real images1
data panel: fMRIPrep's calls14

Select a segment to list its cases below.

Every compared case gives ANTs’ exit status, standard output, standard error and transform file, byte for byte, against the unmodified binary at one thread or, where glibc’s maths is not correctly rounded, against the same binary with correctly rounded maths (CORE-MATH preloaded). Where ANTs never finishes or seeds its sampling from the clock, the replica refuses or is repeatable (expected divergences).

In sMRIPrep: the initial transform of the brain extraction

fMRIPrep’s init_aff, with its arguments, inputs and seed, and niworkflows’ antsAffineInitializer call on the same two images. A transform is shown three ways: the T1w it places on the template (select Checkerboard under Compare to overlay the template and a placed T1w), its parameters, and how far two transforms send the template’s brain apart, in mm.

Worked example

In sMRIPrep: the initial transform of the brain extraction

What this shows: In reproduce, on 1 and 8 threads, the replica of antsAI writes ANTs’ one-thread transform byte for byte, as built and with correctly rounded maths, and so does the replica of antsAffineInitializer with niworkflows’ arguments. fMRIPrep’s transform, ANTs 2.6.2’s at 8 threads, sends the template’s brain at most 2.5e-13 mm from the replica’s, where ANTs 2.6.5’s own 8 and 1 threads differ by 3.2e-13 mm; the placed images are the same bytes. No correction by default changes antsAI’s call; antsAffineInitializer’s default, fix, corrects ANTS-54 and ITK-21 here and moves the brain by up to 0.059 mm (mean 0.054 mm) from reproduce.

fMRIPrep 25.2.6’s init_aff node (antsAI’s multi-start search: rotations around the principal axes, an Affine local search with Mattes MI) on OpenNeuro ds000005 sub-01, with its arguments, inputs and seed; and niworkflows’ antsAffineInitializer call on the same images. Each transform is shown as the T1w it places on the template (resampled by ANTs as built, linear), as its parameters and as distances in mm.

Data
OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01’s T1w (PDDL-1.0); the OASIS30ANTs template (Avants et al., CC BY 4.0, TemplateFlow); shown inside the template’s brain mask, dilated by 2 voxels
Node
fMRIPrep 25.2.6, anat_fit_wf/brain_extraction_wf/init_aff, NSLOTS=8
Command
antsAI -c [10,1e-06,10] -d 3 -x /templateflow/tpl-OASIS30ANTs/tpl-OASIS30ANTs_res-01_desc-BrainCerebellumExtraction_mask.nii.gz -m Mattes[…/res_tmpl/tpl-OASIS30ANTs_res-01_T1w_regrid.nii.gz,…/res_target/sub-01_T1w_noise_corrected_maths_corrected_regrid.nii.gz,32,Regular,0.25] -o initialization.mat -p 0 -s [15,0.1] -g [40.0,0x40x40] -t Affine[0.1] -v 1
Also
antsAffineInitializer 3 <fixed> <moving> out.mat 15 0.1 0 10 (niworkflows’ arguments; fMRIPrep 25.2.6 does not call it here)
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_RANDOM_SEED=50495 (fMRIPrep’s, from the node’s report); the replica at commit 1dfe56c0dc60
Grid
54 × 73 × 64 voxels of 4 × 4 × 4 mm; shown: 54 × 71 × 61 around the brain

Data shown

  • Mixed-gambles task (OpenNeuro ds000005, version 00001), Open Data Commons Public Domain Dedication and Licence 1.0.
  • OASIS30ANTs template (TemplateFlow, version tpl-OASIS30ANTs git e37b3f4e0e50b6c2f79e9e01e8f1aeffcb69fc7f), Creative Commons Attribution 4.0 (CC BY 4.0). Attribution: Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. OASIS30 - Generated with ANTs for MICCAI 2016. https://doi.org/10.1016/j.neuroimage.2010.09.025; https://doi.org/10.1162/jocn.2007.19.9.1498; https://figshare.com/articles/ANTs_ANTsR_Brain_Templates/915436. Distributed by TemplateFlow: Ciric R, et al. TemplateFlow: FAIR-sharing of multi-scale, multi-species brain models. Nat Methods 19, 1568-1571 (2022). doi:10.1038/s41592-022-01681-2 The images shown are adapted from it: reoriented to RAS, cropped to the region shown, masked outside it, and shown as the outputs of the programs compared.
InputOASIS30ANTs template (fixed)as fMRIPrep resampled it for this call
Inputds000005 sub-01 T1w (moving), unalignedN4-corrected, resampled by fMRIPrep; × its brain mask for display; no transform
fMRIPrep's outputfMRIPrep's outputfMRIPrep 25.2.6, as it ran: the T1w placed by fMRIPrep's transform (ANTs 2.6.2, 8 threads)
Replicareplicalarmorx, Apache-2.0 · the T1w placed by its transform; reproduce, 8 threads · against ANTs antsAI 2.6.5

Bit-identical252,288 of 252,288 voxels equal

the original against itself: 1 and 8 threadsANTs antsAI 2.6.5, 8 threads − 1 thread (placed T1w)

Bit-identical252,288 of 252,288 voxels equal

the replica against the originalreplica − ANTs antsAI 2.6.5, 1 thread (placed T1w)

Bit-identical252,288 of 252,288 voxels equal

the replica against fMRIPrep's outputreplica − fMRIPrep's output (placed T1w)

Axial slice 45 of 61; the template and the T1w before and after each transform; Checkerboard (under Compare) overlays the template and a placed T1w. 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 antsAI 2.6.5, 8 threads − 1 thread (placed T1w)referenceThe same placed image.identical: 0 of 252,288 values differ
fMRIPrep's output − ANTs antsAI 2.6.5, 8 threads (placed T1w)referenceThe same placed image.identical: 0 of 252,288 values differ
The replica against the original and fMRIPrep's output
replica − ANTs antsAI 2.6.5, 1 thread (placed T1w)Bit-identicalThe same transform, byte for byte (the .mat file), so the same placed image.identical: 0 of 252,288 values differ
replica − ANTs antsAffineInitializer 2.6.5 (placed T1w)Bit-identicalThe same transform, byte for byte (the .mat file), so the same placed image.identical: 0 of 252,288 values differ
replica − fMRIPrep's output (placed T1w)Bit-identicalThe placed images are the same bytes. The transforms themselves differ: fMRIPrep’s is ANTs at 8 threads, the replica’s ANTs’ one-thread result; they send the brain at most 2.5e-13 mm apart, ANTs’ own 8 and 1 threads 3.2e-13 mm (the table below), far below what the interpolation sees.identical: 0 of 252,288 values differ

Δ is A − B between the placed images, over every voxel of the 54 × 73 × 64 grid.

Where two transforms send the template’s brain apart (mm), and their largest parameter difference

Comparisonthe .mat fileslargest distance (mm)mean distance (mm)largest |Δ parameter|
replica − ANTs antsAI, 1 threadidentical file000
replica − ANTs antsAI, correctly rounded mathsidentical file000
replica − fMRIPrep’s outputdiffer2.5e−131.72e−131.56e−13
replica, 1 thread − 8 threadsidentical file000
replica, fix − reproduceno correction appliesidentical file000
ANTs antsAI, 8 threads − 1 threaddiffer3.24e−132.41e−132.13e−13
fMRIPrep’s output − ANTs antsAI 2.6.5, 8 threadsdiffer1.76e−131.23e−138.53e−14
replica − ANTs antsAffineInitializeridentical file000
replica − ANTs antsAffineInitializer, correctly rounded mathsidentical file000
replica of antsAffineInitializer, fix − reproduceANTS-54, ITK-21 corrected by defaultdiffer0.05930.05450.032
ANTs antsAffineInitializer, 8 threads − 1 threadidentical file000

Over the 91,797 voxels of the template’s brain (OASIS30ANTs’ BrainCerebellumExtraction mask, the call’s -x) on this grid: each transform maps them into the T1w’s space; the distance is between the two images of each point. The shaded rows are ANTs against itself.

The transforms’ parameters (antsAI’s Affine, as written: ITK’s AffineTransform, LPS)

ParameterfMRIPrep (ANTs 2.6.2, 8 threads)ANTs 2.6.5, 1 threadreplicareplica − fMRIPrep's outputreplica − ANTs, 1 thread
matrix 1,11.04175544764990451.0417554476499041.041755447649904−4.44e−160
matrix 1,2−0.02848817664144569−0.028488176641446185−0.028488176641446185−4.96e−160
matrix 1,30.0040677681709451610.0040677681709452990.0040677681709452991.38e−160
matrix 2,10.0503060421761570840.050306042176156540.05030604217615654−5.41e−160
matrix 2,20.96178581268140980.96178581268141040.96178581268141045.55e−160
matrix 2,30.164806211741839680.16480621174183940.1648062117418394−2.78e−160
matrix 3,1−0.024619177697066955−0.024619177697066837−0.0246191776970668371.18e−160
matrix 3,2−0.16642817039749175−0.16642817039749144−0.166428170397491443.05e−160
matrix 3,30.98004208392791550.98004208392791560.98004208392791561.11e−160
translation x−107.72536159701401−107.72536159701396−107.725361597013965.68e−140
translation y−127.035576807221−127.03557680722099−127.035576807220991.42e−140
translation z111.60969639694264111.60969639694248111.60969639694248−1.56e−130

As the .mat files store them (double precision), printed with the digits that identify each value.

Provenance of this example
  • Data: OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01, PDDL-1.0; OASIS30ANTs (Avants et al.), from TemplateFlow, CC BY 4.0
  • Inputs: fixed (sha256 8265828adf534c2e…); moving (sha256 4fc9d00a945155a9…); fMRIPrep’s transform (sha256 f223fa5f0aade78e…)
  • Runs: figures cache entry ants-ai/fmriprep-ds000005-init-aff and ants-ai/fmriprep-ds000005-init-aff-placed, computed 2026-10-21T15:17:05+00:00 with larmorx 1dfe56c0dc60; 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-ai.json, exported by python -m larmorx_validation evidence ants-ai.

Blob alignment

With -b n, antsAI starts its search from a transform fitted to matched blobs instead of the centres of mass: up to n blobs per image from ANTs’ multi-scale Laplacian-of-Gaussian detector; every fixed and moving blob compared by the correlation of their patches (a sphere of radius 20 voxels, turned onto the gradients’ principal axes); the best-correlated pairs whose radii agree; the pairs whose distances to at least three others are preserved become landmarks, and a rigid (or affine) transform is fitted to them. The replica ports every step, and in reproduce gives ANTs’ blobs, correlations, pairs, landmarks and transform bit for bit. ANTs’ code has defects that the replica’s default corrects (findings): its landmark test compares the lengths of the pairs’ displacements, not the distances between blobs (ANTS-110), so any rotation leaves it no landmark and a transform of NaNs (ANTS-114); once the correlations run out it pairs the first two blobs again and again (ANTS-111); its detector does not always keep the strongest maxima (ANTS-112); ITK’s 2D rigid fit takes −90° near a quarter turn (ITK-60); and it compares the patches in whatever orientation the eigenvectors’ signs give, possibly reflected (ANTS-116). At more than one thread ANTs’ -b crashes (ANTS-60); the replica gives its one-thread result at any thread count.

First the blobs: in fMRIPrep’s T1w of ds000005 both programs find the same 30 blobs, each drawn here as a sphere of its radius around its voxel (ANTs’ from its own code, built with prints: validation/probes/ants-ai-blobs).

Worked example

The blobs ANTs and the replica find in a T1w

What this shows: In reproduce the replica finds ANTs’ 30 blobs bit for bit: each at the same voxel, with the same radius and the same Laplacian value as ANTs’ own code finds them. Its default, fix, corrects ANTS-112 to keep the strongest maxima; here the largest scales fill the list before ANTs’ early threshold can drop one, so both lists are the same.

antsAI -b 30 keeps the 30 strongest maxima of the scale-normalised Laplacian of Gaussian over 53 scales; here they lie at the edges of ventricles, vessels and sulci, with radii from 1.7 to 6 mm. Each is drawn as a filled sphere of its radius around its voxel, over the T1w at half its intensity.

Data
OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01’s T1w (PDDL-1.0), as fMRIPrep 25.2.6 preprocessed it
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’ blobs from the blob probe (validation/probes/ants-ai-blobs, antsAI’s code with prints, built against the same ITK); the replica’s from lx.ants.blob_alignment 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 T1w (fixed)fMRIPrep's preprocessed T1w
OriginalANTs antsAI 2.6.5its blobs · antsAI's code with prints (the probe)
Replicareplicalarmorx, Apache-2.0 · its blobs, reproduce, 6 threads · against ANTs antsAI 2.6.5
Replicareplicalarmorx, Apache-2.0 · its blobs, fix (the default, ANTS-112 corrected) · against ANTs antsAI 2.6.5

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

the replica against the originalreplica, reproduce − ANTs antsAI

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

the replica against the originalreplica, fix − ANTs antsAI

Axial slice 66 of 111; each program's blobs, filled spheres over the fixed T1w at half its intensity (the crosshair at the largest blob this slice cuts). 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 replica against the original
replica, reproduce − ANTs antsAIBit-identicalThe same 30 blobs, each at the same voxel with the same radius and the same Laplacian value, bit for bit, as ANTs’ own code finds them.identical: 0 of 5,898,240 values differ
replica, fix − ANTs antsAIBit-identicalANTS-112’s correction keeps the strongest maxima; here the largest scales fill the list before ANTs’ early threshold can drop one, so both lists are the same. ANTS-112identical: 0 of 5,898,240 values differ

Δ is A − B between the drawn blobs, over every voxel of the 160 × 192 × 192 grid.

Provenance of this example
  • Data: OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01, PDDL-1.0
  • Inputs: fixed: fMRIPrep’s preprocessed T1w (sha256 8e977ad0c954252d…); fMRIPrep’s T1w brain mask (sha256 726851b7736940c8…); moving (sha256 2885195494bbf915…)
  • Runs: figures cache entry ants-ai/blobs-t1w, computed 2026-10-23T10:41:20+00:00; 13 of its 32 runs made again since, last on 2026-10-23T22:13:18+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-ai.json, exported by python -m larmorx_validation evidence ants-ai.

Two copies of that T1w with a known transform show what the pairing and the landmark test do. The first is the same voxels with the origin moved: every blob has an identical partner, yet ANTs compares many of those identical patches reflected or turned, pairs eleven blobs (one wrongly) and keeps ten landmarks; it finds the translation exactly. The replica’s default compares each pair of patches in the orientation that matches best: every identical pair correlates at 1, all thirteen blobs whose patches fit are paired with their true partners and kept, and the copy is placed exactly as the truth places it. The tables list every pair, which program kept it, and how far each transform is from the truth.

Worked example

Blob alignment of a T1w and its shifted copy

What this shows: In reproduce, on 6 threads, the replica writes ANTs’ transform file byte for byte, with the same blobs, correlations, pairs and landmarks as ANTs’ own code. Its default, fix, changes the pairs and landmarks (ANTS-116): it keeps 13 landmarks of 13 pairs, ANTs 10 of 11. Both transforms send the copy where the truth does (4.1e-15 and 1.8e-15 mm off at most), so the placed images are the same.

antsAI’s blob alignment (-b 30) between fMRIPrep’s T1w of ds000005 sub-01 and the same voxels with the origin moved by (6, −4, 3) mm, so the true transform is known. Each program’s pairs and landmarks are listed, and its transform places the copy back on the fixed grid.

Data
OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01’s T1w (PDDL-1.0), as fMRIPrep 25.2.6 preprocessed it; the moving image: the same voxels with the origin moved by (6, −4, 3) mm
Command
antsAI -d 3 -m Mattes[fixed.nii.gz,moving.nii.gz,32,Regular,0.25] -t Rigid[0.1] -b 30 -o t.mat -v 1 (the record’s arguments)
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’ steps from the blob probe (validation/probes/ants-ai-blobs, antsAI’s code with prints, built against the same ITK); 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 T1w (fixed)fMRIPrep's preprocessed T1w
Inputits copy (moving), unalignedthe same voxels with the origin moved by (6, −4, 3) mm; × its brain mask, on the fixed grid with no transform
OriginalANTs antsAI 2.6.51 thread · the copy placed by its transform
Replicareplicalarmorx, Apache-2.0 · the copy placed by its transform; fix (the default, ANTS-110 and ANTS-116 corrected), 6 threads · against ANTs antsAI 2.6.5

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

the original against the known answerANTs antsAI − the true transform

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

the replica against the originalreplica, reproduce − ANTs antsAI, 1 thread

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

the replica against the originalreplica, fix − ANTs antsAI, 1 thread

Axial slice 67 of 111; the fixed T1w, its copy before and after each transform; Checkerboard (under Compare) overlays the fixed T1w and a placed copy. 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 the known answer
ANTs antsAI − the true transformreferenceThe same placed image as the true transform’s.identical: 0 of 5,898,240 values differ
The replica against the original and the known answer
replica, reproduce − ANTs antsAI, 1 threadBit-identicalThe same transform file, byte for byte, and the same blobs, correlations, pairs and landmarks as ANTs’ own code computes them (the probe).identical: 0 of 5,898,240 values differ
replica, fix − ANTs antsAI, 1 threadBit-identicalThe same placed image: both transforms send the copy where the truth does (4.1e-15 and 1.8e-15 mm off at most). ANTS-116, corrected by default, changes the pairs and landmarks: the replica keeps 13 landmarks of 13 pairs and ANTs 10 of 11 (the tables below); the patches are compared in the orientation that matches best (proper rotations, every sign combination of their principal axes), where ANTs may compare identical patches reflected or turned. For a rotation ANTs keeps no landmark (the next example). ANTS-116identical: 0 of 5,898,240 values differ
replica, fix − the true transformreferenceThe same placed image as the true transform’s.identical: 0 of 5,898,240 values differ

Δ is A − B between the placed images, over every voxel of the 160 × 192 × 192 grid.

Where two transforms send the T1w’s brain apart (mm), and how far their matrices differ

Comparisonthe .mat fileslargest distance (mm)mean distance (mm)largest |Δ| in the matrix or translation
replica, reproduce − ANTs antsAI, 1 threadidentical file000
replica, reproduce − ANTs antsAI, correctly rounded mathsidentical file000
replica, 1 thread − 6 threadsidentical file000
replica, fix − ANTs antsAIdiffer4.07e−151.57e−151.78e−15
ANTs − the true transformdiffer1.78e−153.79e−161.78e−15
replica, fix − the true transformdiffer4.07e−151.92e−153.55e−15
ANTs, 1 thread: run 2 − run 1identical file000
ANTs antsAI: correctly rounded maths − as builtidentical file000
ANTs, 2 threadscrashed (signal 11)–––
ANTs, 4 threadscrashed (signal 11)–––
ANTs, 6 threadscrashed (signal 11)–––

Over the 763,153 voxels of fMRIPrep’s brain mask of the fixed T1w: each transform maps them into the moving image; the distance is between the two images of each point. The shaded rows are ANTs against itself: at more than one thread its blob detector writes past its per-thread heaps (ANTS-60).

The pairs: each fixed blob matched to a moving blob, and which pairs became landmarks

Pairfixed blob (voxel)moving blob (voxel)radius (mm)patch correlationoff the true transform (mm)ANTsreplica, fix
157 122 8057 122 802.2910landmarklandmark
263 68 9763 68 974.910landmarklandmark
3101 121 80101 121 802.1310landmarklandmark
463 106 9863 106 983.4610landmarklandmark
521 39 10321 40 1052.810.752.98paired–
656 127 8856 127 882.450.6720landmarklandmark
779 85 7679 85 764.260.3350landmarklandmark
861 66 9661 66 964.90.3150landmarklandmark
996 68 9696 68 964.570.2680landmarklandmark
1054 59 8854 59 885.630.2130landmarklandmark
1179 69 6279 69 626.030.1040landmarklandmark
1221 40 10521 40 1053.0210–landmark
1321 39 10321 39 1032.8110–landmark
1422 40 10822 40 1083.0210–landmark

ANTs’ blobs, correlations, pairs and landmarks from the probe (antsAI’s own code); the replica’s default, fix, from lx.ants.blob_alignment. A pair is a landmark when its distances to at least 3 other pairs are preserved within 10 % (squared); ANTs tests the lengths of the pairs’ displacements instead (ANTS-110), so a rotation leaves it no landmark. The replica compares the patches in the orientation that matches best, ANTs in the one its eigenvectors’ signs give, reflected or not (ANTS-116). ‘Off the true transform’: how far the moving blob is from where the true transform sends the fixed blob’s centre.

The transforms as matrices (ITK’s physical space, LPS mm, centre 0)

Entrytrue transformANTs 2.6.5, 1 threadreplica, reproducereplica, fix
matrix 1,11111
matrix 1,20000
matrix 1,30000
matrix 2,10000
matrix 2,21111
matrix 2,30000
matrix 3,10000
matrix 3,20000
matrix 3,31111
translation x−6−6−6−6
translation y4444
translation z3333

antsAI’s output (an Euler3DTransform for Rigid): the blob transform after its search of rotations about the image axes (with no iterations, the best start point), written as the matrix and translation that map a fixed point to the moving image; the true transform as constructed. A transform of NaNs (no landmark) is left out.

Provenance of this example
  • Data: OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01, PDDL-1.0
  • Inputs: fixed: fMRIPrep’s preprocessed T1w (sha256 8e977ad0c954252d…); fMRIPrep’s T1w brain mask (sha256 726851b7736940c8…); moving (sha256 2885195494bbf915…)
  • Runs: figures cache entry ants-ai/blobs-t1w, computed 2026-10-23T10:41:20+00:00; 13 of its 32 runs made again since, last on 2026-10-23T22:13:18+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-ai.json, exported by python -m larmorx_validation evidence ants-ai.

The second is the T1w rotated by 8 degrees (in millimetres) and shifted: no pair keeps the length of its displacement, ANTs keeps no landmark, and its transform (and the replica’s in reproduce, the same file) is NaN. The replica’s default keeps the pairs whose distances are preserved and finds a rotation close to the truth (the metrics below give the angle and the distance from the true transform).

Worked example

Blob alignment of a T1w and its rotated copy: ANTs finds no landmark

What this shows: ANTs keeps no landmark (ANTS-110) and writes a transform of NaNs (ANTS-114); the replica’s reproduce writes the same file. Its default, fix, keeps the pairs whose distances are preserved and finds a rotation of 8.18° (the truth: 8.00°), which sends the brain at most 1.2 mm (mean 0.62 mm) from where the true transform does.

    The same T1w against a copy rotated by 8° and shifted: ANTs’ distance test keeps no landmark and its transform is NaN; the replica’s corrected test keeps the pairs whose distances are preserved and finds a rotation close to the truth.

    Data
    OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01’s T1w (PDDL-1.0), as fMRIPrep 25.2.6 preprocessed it; the moving image: rotated by 8° (in millimetres) about the z axis through the centre and shifted by (5, −3, 2) voxels (linear interpolation)
    Command
    antsAI -d 3 -m Mattes[fixed.nii.gz,moving.nii.gz,32,Regular,0.25] -t Rigid[0.1] -b 30 -o t.mat -v 1 (the record’s arguments)
    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’ steps from the blob probe (validation/probes/ants-ai-blobs, antsAI’s code with prints, built against the same ITK); the replica at commit e22378b480f4
    Grid
    160 × 192 × 192 voxels of 1 × 1.33 × 1.33 mm; shown: 151 × 125 × 111 around the brain
    ANTs
    no pair keeps the lengths of its displacement (ANTS-110), so no landmark: ANTs’ rigid fit divides 0 by 0 and antsAI writes a transform of NaNs (ANTS-114), which places nothing. The replica’s reproduce writes the same file.
    Inputds000005 sub-01 T1w (fixed)fMRIPrep's preprocessed T1w
    Inputits copy (moving), unalignedrotated by 8° (in millimetres) about the z axis through the centre and shifted by (5, −3, 2) voxels (linear interpolation); × its brain mask, on the fixed grid with no transform
    Inputits copy placed by the true transformthe transform known by construction (antsApplyTransforms as built, linear)
    Replicareplicalarmorx, Apache-2.0 · the copy placed by its transform; fix (the default, ANTS-110 and ANTS-116 corrected), 6 threads
    the replica against the known answerreplica, fix − the true transformscale ±1.07e3

    Axial slice 67 of 111; the fixed T1w, its copy before and after each transform; Checkerboard (under Compare) overlays the fixed T1w and a placed copy. 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 replica against the known answer
    replica, fix − the true transformreferenceA rotation of 8.18° (the truth: 8.00°); it sends the brain at most 1.2 mm (mean 0.62 mm) from where the true transform does.1.07e3803,506 of 5,898,240 (13.6 %)19.40.50.00315

    Δ is A − B between the placed images, over every voxel of the 160 × 192 × 192 grid.

    Where two transforms send the T1w’s brain apart (mm), and how far their matrices differ

    Comparisonthe .mat fileslargest distance (mm)mean distance (mm)largest |Δ| in the matrix or translation
    replica, reproduce − ANTs antsAI, 1 threadidentical fileno transform (NaN)––
    replica, reproduce − ANTs antsAI, correctly rounded mathsidentical fileno transform (NaN)––
    replica, 1 thread − 6 threadsidentical fileno transform (NaN)––
    replica, fix − ANTs antsAIdifferno transform (NaN)––
    ANTs − the true transformdifferno transform (NaN)––
    replica, fix − the true transformdiffer1.20.6230.337
    ANTs, 1 thread: run 2 − run 1identical fileno transform (NaN)––
    ANTs antsAI: correctly rounded maths − as builtidentical fileno transform (NaN)––
    ANTs, 2 threadscrashed (signal 11)–––
    ANTs, 4 threadscrashed (signal 11)–––
    ANTs, 6 threadscrashed (signal 11)–––

    Over the 763,153 voxels of fMRIPrep’s brain mask of the fixed T1w: each transform maps them into the moving image; the distance is between the two images of each point. The shaded rows are ANTs against itself: at more than one thread its blob detector writes past its per-thread heaps (ANTS-60).

    The pairs: each fixed blob matched to a moving blob, and which pairs became landmarks

    Pairfixed blob (voxel)moving blob (voxel)radius (mm)patch correlationoff the true transform (mm)ANTsreplica, fix
    156 127 8858 131 852.450.7442.58pairedlandmark
    257 122 8058 128 782.290.7430.613pairedlandmark
    354 59 8843 66 865.630.6860.659pairedlandmark
    463 68 9753 73 954.90.6110.924paired–
    5101 121 8045 143 912.130.52165.1paired–
    663 106 9861 111 963.460.3810.323pairedlandmark
    721 40 10582 102 753.020.2109paired–
    879 69 6269 73 606.030.880.741–landmark
    9101 121 80101 122 782.130.680.128–landmark
    1061 66 9653 73 954.90.6692.75–landmark
    1196 68 9686 70 944.570.5230.346–landmark

    ANTs’ blobs, correlations, pairs and landmarks from the probe (antsAI’s own code); the replica’s default, fix, from lx.ants.blob_alignment. A pair is a landmark when its distances to at least 3 other pairs are preserved within 10 % (squared); ANTs tests the lengths of the pairs’ displacements instead (ANTS-110), so a rotation leaves it no landmark. The replica compares the patches in the orientation that matches best, ANTs in the one its eigenvectors’ signs give, reflected or not (ANTS-116). ‘Off the true transform’: how far the moving blob is from where the true transform sends the fixed blob’s centre.

    The transforms as matrices (ITK’s physical space, LPS mm, centre 0)

    Entrytrue transformreplica, fix
    matrix 1,10.9902680.989856
    matrix 1,20.1391730.141677
    matrix 1,30−0.0105937
    matrix 2,1−0.139173−0.141766
    matrix 2,20.9902680.989866
    matrix 2,30−0.00817907
    matrix 3,100.00932756
    matrix 3,200.00959793
    matrix 3,310.99991
    translation x3.620593.28409
    translation y−4.39303−4.58051
    translation z−2.66667−2.85837

    antsAI’s output (an Euler3DTransform for Rigid): the blob transform after its search of rotations about the image axes (with no iterations, the best start point), written as the matrix and translation that map a fixed point to the moving image; the true transform as constructed. A transform of NaNs (no landmark) is left out.

    Provenance of this example
    • Data: OpenNeuro ds000005 (Tom et al., 2007), snapshot 00001, sub-01, PDDL-1.0
    • Inputs: fixed: fMRIPrep’s preprocessed T1w (sha256 8e977ad0c954252d…); fMRIPrep’s T1w brain mask (sha256 726851b7736940c8…); moving (sha256 2885195494bbf915…)
    • Runs: figures cache entry ants-ai/blobs-t1w, computed 2026-10-23T10:41:20+00:00; 13 of its 32 runs made again since, last on 2026-10-23T22:13:18+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-ai.json, exported by python -m larmorx_validation evidence ants-ai.

    ANTS-116, corrected by default. The patch comparison turns the moving patch onto the fixed one with a rotation built from the patches’ principal axes; ANTs never checks that rotation for a reflection, and the axes’ signs are whatever the eigensolver returns, so identical patches can be compared reflected or turned by 180°. Where the input makes sense and the result is wrong, the replica corrects it: it tries every sign combination of the axes (four in 3D, two in 2D), each with a proper rotation, and keeps the best correlation; ANTs’ own rotation is one of the candidates when it is proper, and reproduce keeps ANTs’ method. Measured on the T1w above and on blob phantoms with a known transform (investigation):

    ANTS-116, corrected by default: the blob patches compared in the orientation that matches best, against ANTs’ method (the eigenvectors’ signs as they come, reflections included); antsAI -b 30, rigid

    Moving imageidentical pairs at 1true / matched pairstrue / landmarksmean distance (mm)largest distance (mm)
    fMRIPrep’s T1w of ds000005, its origin movedANTs’ method (reproduce)4 of 1310 / 1110 / 1000
    The replica13 of 1313 / 1313 / 1300
    the same T1w rotated by 8° and shiftedANTs’ method (reproduce)–5 / 70 / 0no landmark–
    The replica–7 / 87 / 81.081.95
    3D blob phantom, a translationANTs’ method (reproduce)5 of 1924 / 2624 / 2400
    The replica19 of 1919 / 1919 / 1900
    3D blob phantom, rotated by 7° and 4° and shiftedANTs’ method (reproduce)–15 / 1613 / 133466.8
    The replica–9 / 119 / 90.4950.647
    2D blob phantom, a translationANTs’ method (reproduce)14 of 2819 / 2019 / 1900
    The replica28 of 2828 / 2828 / 2800
    2D blob phantom, rotated by 9° and shiftedANTs’ method (reproduce)–9 / 176 / 76.212.3
    The replica–10 / 1610 / 110.851.24

    From validation/investigations/ants-116-blob-orientation/results.json (measure.py: lx.ants.blob_alignment at 30 blobs in both modes; the replica’s default, fix, also corrects ANTS-110, -111, -112, -114 and ITK-60, reproduce is ANTs’ result bit for bit). Identical pairs: the blobs found at the same voxel in both images whose patches fit in both; a pair is true when the moving blob lies within two voxels of where the known transform sends the fixed one; the distances are those of the brain’s (or the phantom’s) points from where the true transform sends them.

    Between subjects, or to a template, the patch matching rarely pairs the same anatomy: on the five fMRIPrep runs’ T1w against MNI152NLin2009cAsym (2 mm), neither program keeps a landmark with 30 blobs (the full comparison’s real/*/blobs-mni-rigid cases). The blob alignment is a rough initialiser for images of one subject.

    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. Each record of the data panel replays every call fMRIPrep 25.2.6 made of the program in its fourteen runs, 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 explorers below (categories data panel: fMRIPrep’s calls and data panel: other images).

    antsAI: 40 of 40 identical to ANTs at one thread (38 as built, 2 with correctly rounded maths): all 36 calls of the fourteen runs (the brain extraction’s initial transform in every run, the BOLD reference’s in 22) with fMRIPrep’s seed, and four images fMRIPrep never gives it. antsAffineInitializer: 14 of 14 identical to ANTs as built, niworkflows’ call inside every run’s template registration, on its inputs remade exactly. Against fMRIPrep’s own transforms (ANTs 2.6.2 at 6 or 8 threads), 31 of the 36 antsAI calls are within 1e-8 in a parameter; five are further, up to 6.8 mm over a 100 mm cube. They are ANTs’ own spread above one thread: fMRIPrep’s ANTs at one thread gives exactly the replica’s transform, and at fMRIPrep’s thread count both ANTs builds land, run to run, on fMRIPrep’s file, on the one-thread result or elsewhere.

    fMRIPrep’s antsAI 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/brain_extraction_wf/init_aff141414014 pass–
    bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff221222022 pass–

    From the panel record (docs/validation/ants-ai-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 brain-extraction antsAI on a brain with a tumour

    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 (sMRIPrep’s init_aff: OASIS30ANTs regridded to 4 mm against the N4-corrected T1w regridded alike, Mattes, an Affine search from fMRIPrep’s seed) on ds005003 sub-RS001: each run’s transform places the T1w (4 mm) on the template’s grid (ANTs’ antsApplyTransforms, linear). A key dataset, shown as a whole volume; every run’s parameters are in the table below it.

    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/init_aff of the fMRIPrep run ds005003-sub-RS001-noreconall
    Command
    antsAI -c [10,1e-06,10] -d 3 -x tpl-OASIS30ANTs_res-01_desc-BrainCerebellumExtraction_mask.nii.gz -m Mattes[tpl-OASIS30ANTs_res-01_T1w_regrid.nii.gz,sub-RS001_ses-1_T1w_noise_corrected_maths_corrected_regrid.nii.gz,32,Regular,0.25] -o initialization.mat -p 0 -s [15,0.1] -g [40.0,0x40x40] -t Affine[0.1] -v 1
    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 349d3f781431
    Grid
    54 × 73 × 64 voxels of 4 × 4 × 4 mm; shown 54 × 73 × 64, every slice

    Data shown

    • Retrospective Task/Rest fMRI Data from Tumor Patients (OpenNeuro ds005003, version 2.0.0, doi:10.18112/openneuro.ds005003.v2.0.0), CC0 1.0 public-domain dedication.
    • OASIS30ANTs template (TemplateFlow, version tpl-OASIS30ANTs git e37b3f4e0e50b6c2f79e9e01e8f1aeffcb69fc7f), Creative Commons Attribution 4.0 (CC BY 4.0). Attribution: Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. OASIS30 - Generated with ANTs for MICCAI 2016. https://doi.org/10.1016/j.neuroimage.2010.09.025; https://doi.org/10.1162/jocn.2007.19.9.1498; https://figshare.com/articles/ANTs_ANTsR_Brain_Templates/915436. Distributed by TemplateFlow: Ciric R, et al. TemplateFlow: FAIR-sharing of multi-scale, multi-species brain models. Nat Methods 19, 1568-1571 (2022). doi:10.1038/s41592-022-01681-2 The images shown are adapted from it: reoriented to RAS, cropped to the region shown, masked outside it, and shown as the outputs of the programs compared.
    InputOASIS30ANTs at 4 mm (fMRIPrep's res_tmpl)tpl-OASIS30ANTs_res-01_T1w_regrid.nii.gz
    OriginalANTs antsAI 2.6.51 thread · as built (GCC 14.3), glibc's maths
    Replicareplicalarmorx, Apache-2.0 · reproduce, 8 threads · against ANTs antsAI 2.6.5
    Replicareplicalarmorx, Apache-2.0 · fix (the default), 8 threads

    Bit-identical252,288 of 252,288 voxels equal

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

    Bit-identical252,288 of 252,288 voxels equal

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

    Bit-identical252,288 of 252,288 voxels equal

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

    Axial slice 33 of 64; 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 antsAI 2.6.5, 8 threads − 1 threadreferenceANTs’ own result moves with its thread count (placed): the yardstick for any difference above.1.12e−5137,974 of 252,288 (54.7 %)7.95e−71.61e−82.2e−17
    The replica against the original
    replica − ANTs antsAI 2.6.5, 1 thread, as builtBit-identicalThe replica gives the unmodified ANTs program’s bytes.identical: 0 of 252,288 values differ
    replica − ANTs antsAI 2.6.5, correctly rounded mathsBit-identicalThe replica gives ANTs’ bytes with correctly rounded maths.identical: 0 of 252,288 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 252,288 values differ
    replica, 1 thread − 8 threads (reproduce)Bit-identicalThe replica’s result does not depend on its thread count.identical: 0 of 252,288 values differ

    Δ is A − B over all 252,288 values of each output (the moving image placed on the fixed grid through each run’s transform).

    Provenance of this example
    • Data: OpenNeuro ds005003, sub-RS001, CC0-1.0
    • Inputs: input 1 (sha256 8265828adf534c2e…)
    • Runs: figures cache entry ants-ai/panel-fmriprep-init-aff-tumour, computed 2026-10-26T21:59:46+00:00 with larmorx 349d3f781431; 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-ai.json, exported by python -m larmorx_validation evidence ants-ai.

    Data panel: fMRIPrep’s brain-extraction antsAI on a brain with a tumour: every run’s transform

    RunClassp0p1p2p3p4p5p6p7p8p9p10p11Centre (fixed parameters)
    ANTs antsAI 2.6.51 thread · as built (GCC 14.3), glibc’s mathsAffineTransform_double_3_31.0587728028066960.19171048733797263−0.04286202631986694−0.237662911491509951.04827554700829270.107607403420840910.05963445644369763−0.195628872071647171.019417726235834−113.17549257498075−167.87445482743215134.00819334183782109.387505 156.142776 -127.291084
    ANTs antsAI 2.6.5, correctly rounded maths1 thread, CORE-MATH preloadedAffineTransform_double_3_31.0587728028066960.19171048733797263−0.04286202631986694−0.237662911491509951.04827554700829270.107607403420840910.05963445644369763−0.195628872071647171.019417726235834−113.17549257498075−167.87445482743215134.00819334183782109.387505 156.142776 -127.291084
    ANTs antsAI 2.6.58 threads · as builtAffineTransform_double_3_31.05877280316980980.19171048736603305−0.042862026454130495−0.237662911605300921.04827554714047140.107607403319016720.05963445650283639−0.195628872111695441.0194177262495052−113.17549258143282−167.8744547821187134.00819334157342109.387505 156.142776 -127.291084
    replicalarmorx, Apache-2.0 · reproduce, 8 threadsAffineTransform_double_3_31.0587728028066960.19171048733797263−0.04286202631986694−0.237662911491509951.04827554700829270.107607403420840910.05963445644369763−0.195628872071647171.019417726235834−113.17549257498075−167.87445482743215134.00819334183782109.387505 156.142776 -127.291084
    replicalarmorx, Apache-2.0 · reproduce, 1 threadAffineTransform_double_3_31.0587728028066960.19171048733797263−0.04286202631986694−0.237662911491509951.04827554700829270.107607403420840910.05963445644369763−0.195628872071647171.019417726235834−113.17549257498075−167.87445482743215134.00819334183782109.387505 156.142776 -127.291084
    replicalarmorx, Apache-2.0 · fix (the default), 8 threadsAffineTransform_double_3_31.0587728028066960.19171048733797263−0.04286202631986694−0.237662911491509951.04827554700829270.107607403420840910.05963445644369763−0.195628872071647171.019417726235834−113.17549257498075−167.87445482743215134.00819334183782109.387505 156.142776 -127.291084

    Every parameter as stored (the shortest float64 that reads back exactly).

    Worked example

    Data panel: sMRIPrep's antsAffineInitializer on a brain with a stroke

    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-54 and ITK-21 here.

    The antsAffineInitializer call niworkflows’ SpatialNormalization makes inside fMRIPrep’s template registration of ds004884 sub-M2079 (no node of its own): the masked 1 mm MNI template and the masked T1w, made again exactly from the node’s inputs; each run’s transform places the T1w on the template’s grid. Shown as three slices.

    Data
    OpenNeuro ds004884, sub-M2079 (CC0-1.0); the whole field of view shown, only margins where every array is 0 left out
    Node
    inside anat_fit/register_template[MNI152NLin2009cAsym]/registration of the fMRIPrep run ds004884-sub-M2079-noreconall
    Command
    antsAffineInitializer 3 fixed_masked.nii.gz moving_masked.nii.gz transform.mat 15.000000 0.100000 0 10
    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 1fa019a08fbb
    Grid
    193 × 229 × 193 voxels of 1 × 1 × 1 mm; shown 193 × 229 × 193, three full-resolution slices

    Data shown

    InputMNI152NLin2009cAsym 1 mm, masked (niworkflows')fixed_masked.nii.gz
    OriginalANTs antsAffineInitializer 2.6.51 thread · as built (GCC 14.3), glibc's maths
    Replicareplicalarmorx, Apache-2.0 · reproduce, 8 threads · against ANTs antsAffineInitializer 2.6.5
    Replicareplicalarmorx, Apache-2.0 · fix (the default), 8 threads

    Bit-identical8,530,021 of 8,530,021 voxels equal

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

    Bit-identical8,530,021 of 8,530,021 voxels equal

    the replica against the original: correctly rounded mathsreplica − ANTs antsAffineInitializer 2.6.5, correctly rounded maths
    the replica against itself: fix and reproducereplica, fix − reproducescale ±1.07e3

    Axial slice 97 of 193; 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 (97, 115, 97): 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 reasonmax |Δ|Values that differRMS of Δmax |Δ| ÷ range1 − NCC
    The original against itself: the yardstick
    ANTs antsAffineInitializer 2.6.5, 8 threads − 1 threadreferenceANTs’ own result moves with its thread count (placed): the yardstick for any difference above.1.05e32,062,593 of 8,530,021 (24.2 %)50.10.8740.012
    The replica against the original
    replica − ANTs antsAffineInitializer 2.6.5, 1 thread, as builtBit-identicalThe replica gives the unmodified ANTs program’s bytes.identical: 0 of 8,530,021 values differ
    replica − ANTs antsAffineInitializer 2.6.5, correctly rounded mathsBit-identicalThe replica gives ANTs’ bytes with correctly rounded maths.identical: 0 of 8,530,021 values differ
    The replica against itself: thread counts, fix and reproduce
    replica, fix − reproduceDeliberate correctionThe replica’s default, fix, corrects ANTS-54, ITK-21 (register); reproduce gives ANTs’ result. Changed: placed.1.2e32,048,939 of 8,530,021 (24 %)17710.154
    replica, 1 thread − 8 threads (reproduce)Bit-identicalThe replica’s result does not depend on its thread count.identical: 0 of 8,530,021 values differ

    Δ is A − B over all 8,530,021 values of each output (the moving image placed on the fixed grid through each run’s transform).

    Provenance of this example
    • Data: OpenNeuro ds004884, sub-M2079, CC0-1.0
    • Inputs: input 1 (sha256 da7037c3cb139408…)
    • Runs: figures cache entry ants-ai/panel-affine-initializer-stroke, computed 2026-10-26T23:21:01+00:00 with larmorx 1fa019a08fbb; 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-ai.json, exported by python -m larmorx_validation evidence ants-ai.

    Data panel: sMRIPrep’s antsAffineInitializer on a brain with a stroke: every run’s transform

    RunClassp0p1p2p3p4p5p6p7p8p9p10p11Centre (fixed parameters)
    ANTs antsAffineInitializer 2.6.51 thread · as built (GCC 14.3), glibc’s mathsAffineTransform_double_3_30.9025548923126273−6.826296622480044e−40.045978362309406970.0066838124374407110.91577837546442040.31659346904587743−0.03804590103957365−0.302118875865421350.824649006148793−5.002696301399927−23.985073977298526−8.3244901420048480.235927808 21.9994855 9.74520071
    ANTs antsAffineInitializer 2.6.5, correctly rounded maths1 thread, CORE-MATH preloadedAffineTransform_double_3_30.9025548923126273−6.826296622480044e−40.045978362309406970.0066838124374407110.91577837546442040.31659346904587743−0.03804590103957365−0.302118875865421350.824649006148793−5.002696301399927−23.985073977298526−8.3244901420048480.235927808 21.9994855 9.74520071
    ANTs antsAffineInitializer 2.6.58 threads · as builtAffineTransform_double_3_30.90196911221252120.00166646199310652660.050362039853679640.0028272776489175710.92185050732871670.28720929261967193−0.04284155617296085−0.28146828923938320.834196498218286−5.09028804738991−24.21276850861694−8.1442998514345460.235927808 21.9994855 9.74520071
    replicalarmorx, Apache-2.0 · reproduce, 8 threadsAffineTransform_double_3_30.9025548923126273−6.826296622480044e−40.045978362309406970.0066838124374407110.91577837546442040.31659346904587743−0.03804590103957365−0.302118875865421350.824649006148793−5.002696301399927−23.985073977298526−8.3244901420048480.235927808 21.9994855 9.74520071
    replicalarmorx, Apache-2.0 · reproduce, 1 threadAffineTransform_double_3_30.9025548923126273−6.826296622480044e−40.045978362309406970.0066838124374407110.91577837546442040.31659346904587743−0.03804590103957365−0.302118875865421350.824649006148793−5.002696301399927−23.985073977298526−8.3244901420048480.235927808 21.9994855 9.74520071
    replicalarmorx, Apache-2.0 · fix (the default), 8 threadsAffineTransform_double_3_30.9938226635705839−0.0449548745488160750.0058627046781428590.042826990459596860.97468716006850730.20974342635270107−0.0156843538637004−0.209377776816340540.9744147948752606−6.355562493691748−23.920909807336287−8.8442811974335860.235927808 21.9994855 9.74520071

    Every parameter as stored (the shortest float64 that reads back exactly).

    Every case

    Each case of the two programs’ records, with what it tests, its verdict and the reason. Search, filter by category or verdict, sort by a column; #case=<program>/<name> links to one.

    antsAI

    172 cases

    CaseCategoryWhat it testsVerdictChecks
    antsAI/initial/rigidinitial alignmentsRigid, no search: the centres of massBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/initial/rigid-principal-axesinitial alignmentsRigid from the principal axes (Wahba)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/initial/affine-principal-axesinitial alignmentsAffine from the principal axesBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/initial/similarity-principal-axesinitial alignmentsSimilarity from the principal axes (Similarity3D's matrix: cbrt)Identical with correctly rounded maths

    differs from ANTs as built in t.mat; identical to ANTs with CORE-MATH preloaded

    11/11
    antsAI/initial/geometric-centersinitial alignmentsAlignGeometricCentersBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/initial/centers-of-massinitial alignmentsAlignCentersOfMassBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/initial/2d-rigid-principal-axesinitial alignments2D Rigid from the principal axesBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/initial/2d-centers-of-mass-txtinitial alignments2D AlignCentersOfMass, a text fileBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/initial/h5initial alignmentsAffine from the principal axes into an HDF5 fileBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/rigid-gridthe multi-start searchRigid search with a translation gridBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAI/search/similaritythe multi-start searchSimilarity searchBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/affine-mithe multi-start searchAffine search, joint-histogram MIBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAI/search/rigid-gcthe multi-start searchRigid search from the principal axes, GCBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAI/search/affine-obliquethe multi-start searchAffine search on oblique grids (1.5 x 1.2 x 2 mm)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/masksthe multi-start searchAffine search with fixed and moving masksBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/fixed-maskthe multi-start searchRigid search with a fixed maskBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/missing-maskthe multi-start searchA mask that cannot be read is no maskBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/densethe multi-start searchRigid search over every voxel (no sampling)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/gc-densethe multi-start searchGC over every voxelBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/2d-affine-gridthe multi-start search2D Affine search with a translation gridBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAI/search/2d-rigid-gridthe multi-start search2D Rigid with a translation grid (ANTS-52: the grid is ignored)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/2d-rigidthe multi-start search2D Rigid from the principal axesBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/2d-similarity-gcthe multi-start search2D Similarity, GC over every pixelBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/2d-mi-densethe multi-start search2D Affine, MI over every pixelBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/similarity-gridthe multi-start search3D Similarity with a translation grid (ANTS-52: the grid is ignored)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/grid-two-extentsthe multi-start searchA grid of two extents in 3D (ANTS-55: the third is read past the end)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/grid-one-extentthe multi-start searchA grid of one extent in 3D (ANTS-55)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/arc-zerothe multi-start searchAn arc fraction of 0: one angle per axisBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/arc-widethe multi-start searchAn arc fraction of 3 (beyond a full turn)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/seedthe multi-start search--random-seed 7Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/env-seedthe multi-start searchANTS_RANDOM_SEED=99Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/env-seed-prefixthe multi-start searchANTS_RANDOM_SEED=12abc (std::stoi reads 12)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/percentage-above-onethe multi-start searchRegular sampling of 5 (every voxel)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/quietthe multi-start searchWithout -v: nothing printedBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/txtthe multi-start searchA text transform fileBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/h5the multi-start searchAn HDF5 transform file (no compression after a search)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/plain-valuesthe multi-start search-s 90 and -c 2 without bracketsBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/two-outputsthe multi-start searchTwo -o: the last is writtenBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/verbose-lastthe multi-start search-v 1 -v 0: the last one countsBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/unknown-metric-initialthe multi-start searchAn unknown metric with 0 iterations: the metric is not usedBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/empty-start-list-no-outputthe multi-start searchA negative arc fraction without -o: no start point, nothing writtenBit-identical

    exit status, standard output, standard error, every file

    5/5
    antsAI/search/align-with-iterationsthe multi-start searchAlignGeometricCenters with iterations (ANTS-53: an empty .mat)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/align-with-iterations-quietthe multi-start searchAlignGeometricCenters with iterations, without -v (ANTS-53)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/align-with-iterations-2d-gcthe multi-start search2D AlignCentersOfMass with iterations, GC (ANTS-53)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/align-with-iterations-no-startthe multi-start searchAlignGeometricCenters with iterations and a negative arc (ANTS-53)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/gc-empty-mask-densethe multi-start searchGC over every voxel with an empty fixed mask: no valid point, the maximum valueBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/search/mi-empty-mask-densethe multi-start searchMI over every voxel with an empty fixed mask (ITK-19: every start point keeps the first value)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/messages/no-argumentsmessagesNo arguments: the short menu, exit 1Rejected by both

    Print the help menu.

    2/2
    antsAI/messages/hmessages-h: the short menuBit-identical

    exit status, standard output, standard error, every file

    5/5
    antsAI/messages/helpmessages--help: the long menuBit-identical

    exit status, standard output, standard error, every file

    5/5
    antsAI/messages/help-0messages--help 0: nothing to do (ANTS-45)Rejected by both

    Image dimensionality not specified. See command line option --dimensionality

    2/2
    antsAI/messages/h-0messages-h 0 (ANTS-45)Rejected by both

    Image dimensionality not specified. See command line option --dimensionality

    2/2
    antsAI/messages/no-dimensionmessagesNo -dRejected by both

    Image dimensionality not specified. See command line option --dimensionality

    2/2
    antsAI/messages/dimension-4messages-d 4Rejected by both

    Unrecognized dimensionality. Please see help menu.

    2/2
    antsAI/messages/dimension-1messages-d 1Rejected by both

    Unrecognized dimensionality. Please see help menu.

    2/2
    antsAI/messages/no-metricmessagesNo -mRejected by both

    Input metric not specified.

    2/2
    antsAI/messages/no-metric-quietmessagesNo -m, without -vRejected by both

    Input metric not specified.

    1/2
    antsAI/messages/one-imagemessages-m with one imageRejected by both

    Fixed and/or moving images not specified.

    2/2
    antsAI/messages/unknown-samplingmessagesAn unknown samplingRejected by both

    Unrecognized sampling strategy.

    2/2
    antsAI/messages/unknown-transformmessagesAn unknown transformRejected by both

    Unrecognized transform option.

    2/2
    antsAI/messages/unknown-transform-quietmessagesAn unknown transform, without -v (ANTS-45)Rejected by both

    Unrecognized transform option.

    2/2
    antsAI/messages/no-transformmessagesNo -tRejected by both

    Unrecognized transform option.

    2/2
    antsAI/messages/unknown-metricmessagesAn unknown metric with iterationsRejected by both

    ERROR: Unrecognized metric.

    2/2
    antsAI/messages/unknown-metric-alignmessagesAn unknown metric, AlignCentersOfMass with iterationsRejected by both

    ERROR: Unrecognized metric.

    2/2
    antsAI/messages/two-blobsmessages-b 2: too few blobsRejected by both

    Please specify a greater number of blobs (>=3).

    2/2
    antsAI/messages/two-blobs-bracketmessages-b [2,5]Rejected by both

    Please specify a greater number of blobs (>=3).

    2/2
    antsAI/messages/unknown-flagmessagesAn unknown flagRejected by both

    ERROR: Invalid flag provided bogus

    2/2
    antsAI/messages/missing-fixedmessagesA fixed image that does not existRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/missing-moving-geometrymessagesAlignGeometricCenters with a missing moving imageRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/missing-fixed-massmessagesAlignCentersOfMass with a missing fixed imageRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/bad-iterationsmessages-c 1o: the parser's conversion failsRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/bad-binsmessages32 bins written 3x2Rejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/unbalanced-bracketmessagesAn unbalanced bracketRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/bad-seedmessages--random-seed 1xRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/bad-env-seedmessagesANTS_RANDOM_SEED=abc (std::stoi throws)Rejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/env-seed-rangemessagesANTS_RANDOM_SEED out of int's rangeRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/zero-moving-massmessagesAlignCentersOfMass, a moving image of zerosRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/zero-moving-geometrymessagesAlignGeometricCenters, a moving image of zeros: no moments neededBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/messages/zero-principal-axesmessagesPrincipal axes of an image of zerosRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/unknown-extensionmessagesAn output name no writer takesRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/no-directorymessagesAn output in a directory that does not existRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/align-iterations-unknown-extensionmessagesANTS-53 with an output name no writer takesRejected by both

    larmorx: upstream defect ANTS-53 (antsAI) reproduced: AlignGeometricCenters and AlignCentersOfMass with iterations write no transform, as in ANTs [LARMORX_UPSTREAM_DEFECTS=fix corrects it]

    3/3
    antsAI/messages/align-iterations-txtmessagesANTS-53 into a text file: ITK's writer reads an empty listRejected by both

    larmorx: upstream defect ANTS-53 (antsAI) reproduced: AlignGeometricCenters and AlignCentersOfMass with iterations write no transform, as in ANTs [LARMORX_UPSTREAM_DEFECTS=fix corrects it]

    3/3
    antsAI/messages/align-iterations-h5messagesANTS-53 into an HDF5 fileRejected by both

    larmorx: upstream defect ANTS-53 (antsAI) reproduced: AlignGeometricCenters and AlignCentersOfMass with iterations write no transform, as in ANTs [LARMORX_UPSTREAM_DEFECTS=fix corrects it]

    3/3
    antsAI/messages/empty-start-listmessagesA negative arc fraction: no start point, the best of an empty listRejected by both

    larmorx: antsAI: no start point (a negative arc fraction or translation grid): ANTs reads the best parameters from an empty list and crashes

    3/3
    antsAI/messages/empty-gridmessagesA negative grid extent: no start pointRejected by both

    larmorx: antsAI: no start point (a negative arc fraction or translation grid): ANTs reads the best parameters from an empty list and crashes

    3/3
    antsAI/messages/empty-sample-mattesmessagesRegular sampling inside an empty mask: no sample pointRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/empty-sample-mimessagesThe same with MIRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/empty-sample-gcmessagesThe same with GCRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/empty-sample-alignmessagesNo sample point, AlignGeometricCenters with iterationsRejected by both

    larmorx: upstream defect ANTS-53 (antsAI) reproduced: AlignGeometricCenters and AlignCentersOfMass with iterations write no transform, as in ANTs [LARMORX_UPSTREAM_DEFECTS=fix corrects it]

    3/3
    antsAI/messages/random-zeromessagesRandom sampling of 0: no sample pointRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/messages/mattes-empty-mask-densemessagesMattes over every voxel with an empty fixed mask (ITK-20: the second start point crashes)Rejected by both

    larmorx: antsAI: start point 0 threw (Negative spacing is not supported and may result in undefined behavior. Refusing to change spacing from [1, 1] to [-2.43059e+37, 33.1286]); ITK’s Mattes metric is

    3/3
    antsAI/messages/mattes-empty-moving-maskmessagesMattes with an empty moving mask (ITK-20)Rejected by both

    larmorx: antsAI: start point 0 threw (Negative spacing is not supported and may result in undefined behavior. Refusing to change spacing from [1, 1] to [16.229, -2.43059e+37]); ITK’s Mattes metric is

    3/3
    antsAI/messages/mattes-one-start-pointmessagesMattes with an empty mask and one start point: the best value of an empty listRejected by both

    larmorx: antsAI: every start point threw: ANTs’ observer reads the best metric value from an empty list and crashes

    3/3
    antsAI/messages/percentage-zeromessagesRegular sampling of 0 (ANTS-06), one sample point: one intensity (ITK-20)Rejected by both

    larmorx: upstream defect ANTS-06 (ANTs) reproduced: the Regular sampling step: unsigned long of inf is undefined; 9223372036854775808, as x86-64 converts it [LARMORX_UPSTREAM_DEFECTS=fix corrects it]

    3/3
    antsAI/messages/search-factor-negativemessagesA negative search factor: ANTs' loop never endsExpected divergence

    ANTs never finishes (class (a)); the replica refuses

    2/2
    antsAI/messages/translation-step-zeromessagesA translation step of 0: ANTs' loop never endsExpected divergence

    ANTs never finishes (class (a)); the replica refuses

    2/2
    antsAI/random/random-samplingrandom samplingRandom sampling: ANTs draws the voxels from the clock (ANTS-58)Expected divergence

    ANTs seeds from the clock (ANTS-58): two runs differ; the replica is repeatable

    3/3
    antsAI/random/seed-zerorandom sampling--random-seed 0: ANTs seeds Regular sampling from the clock (ANTS-58)Expected divergence

    ANTs seeds from the clock (ANTS-58): two runs differ; the replica is repeatable

    3/3
    antsAI/real/ds000005-sub-01/brain-extraction worked examplereal imagesfMRIPrep's init_aff, brain extraction, with fMRIPrep's seedBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    8/8
    antsAI/real/ds000005-sub-01/bold-mixedgamblestask-run-01real imagesfMRIPrep's init_aff, BOLD reference (mixedgamblestask_run_01), with fMRIPrep's seedBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    8/8
    antsAI/blobs/shift-rigidthe blob alignmentRigid from 30 blobs of a phantom and its copy moved by an exact translation (ANTs' own test keeps the true pairs)Bit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter; from the known transform: ANTs 0 mm mean, 0 mm largest, the replica’s default, fix, 0 mm mean, 0 mm largest

    9/9
    antsAI/blobs/shift-affinethe blob alignmentAffine from 30 blobs of a phantom and its copy moved by an exact translation (ANTs' own test keeps the true pairs)Bit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 1.11e-14 mm mean, 2.2e-14 mm largest, the replica’s default, fix, 6.69e-15 mm mean, 1.04e-14 mm largest

    8/8
    antsAI/blobs/shift-similaritythe blob alignmentSimilarity from 30 blobs of a phantom and its copy moved by an exact translation (ANTs' own test keeps the true pairs)Bit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 0 mm mean, 0 mm largest, the replica’s default, fix, 0 mm mean, 0 mm largest

    8/8
    antsAI/blobs/rigid-rigidthe blob alignmentRigid from 30 blobs of a phantom and its copy moved by a rotation of 7 and 4 degrees and a shiftBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 34 mm mean, 66.8 mm largest, the replica’s default, fix, 0.495 mm mean, 0.647 mm largest

    8/8
    antsAI/blobs/rigid-affinethe blob alignmentAffine from 30 blobs of a phantom and its copy moved by a rotation of 7 and 4 degrees and a shiftBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 654 mm mean, 1.52e+03 mm largest, the replica’s default, fix, 1.41 mm mean, 2.89 mm largest

    8/8
    antsAI/blobs/affine-rigidthe blob alignmentRigid from 30 blobs of a phantom and its copy moved by a rotation of 5 degrees with scales 1.06 and 0.96Bit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs nan mm mean, nan mm largest, the replica’s default, fix, 2.44 mm mean, 5.5 mm largest

    8/8
    antsAI/blobs/affine-affinethe blob alignmentAffine from 30 blobs of a phantom and its copy moved by a rotation of 5 degrees with scales 1.06 and 0.96Rejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/blobs/2d-shift-rigidthe blob alignmentRigid from 30 blobs of a phantom and its copy moved by a 2D translationBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 0 mm mean, 0 mm largest, the replica’s default, fix, 2.66e-15 mm mean, 7.11e-15 mm largest

    8/8
    antsAI/blobs/2d-shift-affinethe blob alignmentAffine from 30 blobs of a phantom and its copy moved by a 2D translationBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 5.81e-14 mm mean, 2.07e-13 mm largest, the replica’s default, fix, 2.61e-14 mm mean, 7.03e-14 mm largest

    8/8
    antsAI/blobs/2d-rigid-rigidthe blob alignmentRigid from 30 blobs of a phantom and its copy moved by a 2D rotation of 9 degrees and a shiftBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 6.2 mm mean, 12.3 mm largest, the replica’s default, fix, 0.85 mm mean, 1.24 mm largest

    8/8
    antsAI/blobs/2d-rigid-affinethe blob alignmentAffine from 30 blobs of a phantom and its copy moved by a 2D rotation of 9 degrees and a shiftBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 45.4 mm mean, 125 mm largest, the replica’s default, fix, 1.44 mm mean, 4.39 mm largest

    8/8
    antsAI/blobs/2d-rigid-similaritythe blob alignmentSimilarity from 30 blobs of a phantom and its copy moved by a 2D rotation of 9 degrees and a shiftBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 6.2 mm mean, 12.3 mm largest, the replica’s default, fix, 0.85 mm mean, 1.24 mm largest

    8/8
    antsAI/blobs/shift-rigid-match-10the blob alignment-b [30,10]: 30 blobs extracted, at most 10 pairsBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 0 mm mean, 0 mm largest, the replica’s default, fix, 8.88e-17 mm mean, 4.44e-16 mm largest

    8/8
    antsAI/blobs/rigid-affine-search-h5the blob alignmentAffine from the blobs of the rotated phantom, a search, an HDF5 outputBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 603 mm mean, 1.25e+03 mm largest, the replica’s default, fix, 4.48 mm mean, 7.33 mm largest

    8/8
    antsAI/blobs/identical-rigidthe blob alignmentAn image against itself: every displacement is 0, so ANTs' ratios are 0/0 (ANTS-110)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/blobs/no-pairs-rigidthe blob alignment-b [5,0]: no pair to match; the rigid fit of no landmark (ANTS-114)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/blobs/no-pairs-affinethe blob alignment-b [5,0] with Affine: too few landmarks, ITK's exceptionRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/blobs/2d-no-pairs-rigidthe blob alignment2D, no pair: ITK's warning (fewer than 2 landmarks), then NaN (ANTS-114)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/blobs/constantthe blob alignmentTwo constant images: every voxel is a maximum of value 0Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/blobs/noise-rigidthe blob alignmentNoise against itself, 48^3: patches barely fitBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/blobs/thinthe blob alignmentAn axis of 3 voxels: the Laplacian of Gaussian refuses it (ITK's exception)Rejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/blobs/missing-fixedthe blob alignmentA fixed image that does not exist: the blob detector's input is not setRejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/blobs/missing-movingthe blob alignmentA moving image that does not exist (after the fixed image's blobs)Rejected by both

    larmorx: antsAI: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAI/blobs/principal-axes-ignoredthe blob alignment-p 1 with -b: the blobs decide the initial transformBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs 0 mm mean, 0 mm largest, the replica’s default, fix, 0 mm mean, 0 mm largest

    8/8
    antsAI/real/ds000005-sub-01/blobs-t1w-shift worked examplereal imagesfMRIPrep's T1w of ds000005 against its copy with the origin moved (6, -4, 3) mmBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter; from the known transform: ANTs 5.92e-16 mm mean, 1.78e-15 mm largest, the replica’s default, fix, 1.78e-15 mm mean, 4.07e-15 mm largest

    9/9
    antsAI/real/ds000005-sub-01/blobs-t1w-rotated worked examplereal imagesfMRIPrep's T1w of ds000005 against its copy rotated by 8 degrees and shiftedBit-identical

    exit status, standard output, standard error, every file; from the known transform: ANTs nan mm mean, nan mm largest, the replica’s default, fix, 1.08 mm mean, 1.95 mm largest

    8/8
    antsAI/real/ds000005-sub-01/blobs-t1w-rotated-affinereal imagesThe same with Affine (ANTs: too few landmarks)Rejected by both

    larmorx: upstream defect ANTS-110 (antsAI) reproduced: 5 of 7 blob pairs preserve their distances to at least 3 others; ANTs’ test of the displacement lengths keeps 0 [LARMORX_UPSTREAM_DEFECTS=fix cor

    3/3
    antsAI/real/ds000005-sub-01/blobs-mni-rigidreal imagesMNI152NLin2009cAsym (2 mm) against ds000005-sub-01's T1w, Rigid from 30 blobsBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/difftest/scales-warnings-before-multistartrandom testingrandom testing's permanent case: antsAI's start points that throw: each start point's scales warnings before its multi-start warning, in ITK's orderBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/panel/fmriprep/ds000005-sub-01/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds000005-sub-01 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    8/8
    antsAI/panel/fmriprep/ds000005-sub-01/bold[task_mixedgamblestask_run_01]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_mixedgamblestask_run_01]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds000005-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    8/8
    antsAI/panel/fmriprep/ds000005-sub-01/bold[task_mixedgamblestask_run_02]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_mixedgamblestask_run_02]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds000005-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds000005-sub-01/bold[task_mixedgamblestask_run_03]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_mixedgamblestask_run_03]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds000005-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds000228-sub-pixar008/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds000228-sub-pixar008 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds000228-sub-pixar008/bold[task_pixar]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_pixar]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds000228-sub-pixar008 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds000256-sub-CTS203/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds000256-sub-CTS203 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds000256-sub-CTS203/bold[task_restbaseline]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_restbaseline]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds000256-sub-CTS203 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds000256-sub-CTS203/bold[task_restfixationfixed]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_restfixationfixed]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds000256-sub-CTS203 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryIdentical with correctly rounded maths

    differs from ANTs as built in initialization.mat; identical to ANTs with CORE-MATH preloaded

    12/12
    antsAI/panel/fmriprep/ds000256-sub-CTS203/bold[task_restfixationnofeedback]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_restfixationnofeedback]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds000256-sub-CTS203 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds001600-sub-1/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds001600-sub-1 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds002422-sub-02/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds002422-sub-02 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds002422-sub-02/bold[task_arithm_run_01]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_arithm_run_01]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds002422-sub-02 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds002422-sub-02/bold[task_arithm_run_02]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_arithm_run_02]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds002422-sub-02 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds002422-sub-02/bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds002422-sub-02 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds002603-sub-01/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds002603-sub-01 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds002603-sub-01/bold[task_wm_acq_Mb4Mm3Tr550]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_wm_acq_Mb4Mm3Tr550]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds002603-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds002603-sub-01/bold[task_wm_acq_SeqMm3Tr2000]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_wm_acq_SeqMm3Tr2000]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds002603-sub-01 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds003592-sub-280/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds003592-sub-280 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds003592-sub-280/bold[ses_1_task_rest_echo_1]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[ses_1_task_rest_echo_1]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds003592-sub-280 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds004173-sub-603309/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds004173-sub-603309 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds004884-sub-M2079/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds004884-sub-M2079 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds004884-sub-M2079/bold[ses_217_task_naming40_acq_epfid2_dir_AP_run_2]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[ses_217_task_naming40_acq_epfid2_dir_AP_run_2]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds004884-sub-M2079 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds004884-sub-M2079/bold[ses_217_task_rest_acq_epfidp2p2_dir_AP_run_9]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[ses_217_task_rest_acq_epfidp2p2_dir_AP_run_9]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds004884-sub-M2079 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds004958-sub-fsm64ps/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds004958-sub-fsm64ps (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds004958-sub-fsm64ps/bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds004958-sub-fsm64ps (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005003-sub-RS001/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds005003-sub-RS001 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005003-sub-RS001/bold[ses_1_task_motor]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[ses_1_task_motor]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds005003-sub-RS001 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005003-sub-RS001/bold[ses_1_task_rest]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[ses_1_task_rest]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds005003-sub-RS001 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005040-sub-002/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds005040-sub-002 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005040-sub-002/bold[task_clip_run_1]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_clip_run_1]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds005040-sub-002 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005454-sub-16/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds005454-sub-16 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005454-sub-16/bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[task_rest]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds005454-sub-16 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005752-sub-ON01016/anat_fit/brain_extraction/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `anat_fit/brain_extraction/init_aff` of the fMRIPrep run ds005752-sub-ON01016 (anat_fit_wf/brain_extraction_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005752-sub-ON01016/bold[ses_01_task_rest_dir_forward]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[ses_01_task_rest_dir_forward]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds005752-sub-ON01016 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/fmriprep/ds005752-sub-ON01016/bold[ses_01_task_rest_dir_reverse]/bold_fit/enhance_and_skullstrip_bold/init_affdata panel: fMRIPrep's callsfMRIPrep 25.2.6's antsAI node `bold[ses_01_task_rest_dir_reverse]/bold_fit/enhance_and_skullstrip_bold/init_aff` of the fMRIPrep run ds005752-sub-ON01016 (bold_wf/bold_fit_wf/enhance_and_skullstrip_bold_wf/init_aff): its command.txt replayed, its inputs from the run's working directoryBit-identical

    exit status, standard output, standard error, every file

    7/7
    antsAI/panel/images/brain-extraction/flairdata panel: other imagessMRIPrep's brain-extraction antsAI (OASIS30ANTs, Mattes, an Affine search, fMRIPrep's seed) on a FLAIR whose 44 slices of 3 mm cover part of the head (ds003592 sub-280: an 82-year-old's FLAIR (0.42 x 0.42 x 3 mm, 44 slices))Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/panel/images/brain-extraction/tumour-t2wdata panel: other imagessMRIPrep's brain-extraction antsAI (OASIS30ANTs, Mattes, an Affine search, fMRIPrep's seed) on a T2w of 6 mm 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))Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/panel/images/brain-extraction/t2w-olderdata panel: other imagessMRIPrep's brain-extraction antsAI (OASIS30ANTs, Mattes, an Affine search, fMRIPrep's seed) on an older adult's T2w (T2 contrast against a T1 template) (ds004958 sub-fsm64ps: 3 T T2-SPACE of a 68-year-old (1 mm, defaced with MiDeFace))Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAI/panel/images/brain-extraction/7t-mp2ragedata panel: other imagessMRIPrep's brain-extraction antsAI (OASIS30ANTs, Mattes, an Affine search, fMRIPrep's seed) on the raw 7 T MP2RAGE, its 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)Identical with correctly rounded maths

    differs from ANTs as built in initialization.mat; identical to ANTs with CORE-MATH preloaded

    11/11

    antsAffineInitializer

    55 cases

    CaseCategoryWhat it testsVerdictChecks
    antsAffineInitializer/messages/no-argumentsmessagesNo arguments: the usage, exit 0Bit-identical

    exit status, standard output, standard error, every file

    5/5
    antsAffineInitializer/messages/dimension-onlymessagesOne argument: the usage, exit 0Bit-identical

    exit status, standard output, standard error, every file

    5/5
    antsAffineInitializer/messages/two-argumentsmessagesTwo arguments: std::string of a null pointer (ANTS-57)Rejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/three-argumentsmessagesThree arguments (ANTS-57)Rejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/dimension-textmessagesA dimension std::stoi cannot readRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/dimension-rangemessagesA dimension beyond int's rangeRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/dimension-5messagesDimension 5 runs as 2D (ANTS-56)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/messages/dimension-1messagesDimension 1 runs as 2D (ANTS-56)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/messages/principal-axes-textmessagesA principal-axes flag std::stoi cannot readRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/missing-fixedmessagesA fixed image that does not existRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/missing-movingmessagesA moving image that does not existRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/zero-fixedmessagesA fixed image of zerosRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/zero-movingmessagesA moving image of zerosRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/unknown-extensionmessagesAn output name no writer takesRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/no-directorymessagesAn output in a directory that does not existRejected by both

    larmorx: antsAffineInitializer: ANTs aborts here (an uncaught exception); the replica stops with exit 1

    3/3
    antsAffineInitializer/messages/search-factor-zeromessagesA search factor of 0: ANTs' loop never endsExpected divergence

    ANTs never finishes (class (a)); the replica refuses

    2/2
    antsAffineInitializer/messages/search-factor-textmessagesA search factor atof reads as 0: ANTs' loop never endsExpected divergence

    ANTs never finishes (class (a)); the replica refuses

    2/2
    antsAffineInitializer/messages/search-factor-nanmessagesA search factor that is not a number: no start pointRejected by both

    larmorx: antsAffineInitializer: no start point (a search factor or radian fraction that is not a number): ANTs reads the best parameters from an empty list and crashes

    4/4
    antsAffineInitializer/messages/radian-fraction-nanmessagesA radian fraction that is not a number: no start pointRejected by both

    larmorx: antsAffineInitializer: no start point (a search factor or radian fraction that is not a number): ANTs reads the best parameters from an empty list and crashes

    4/4
    antsAffineInitializer/search/defaultsthe multi-start search3D, every default (10 degrees, pi/4)Bit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAffineInitializer/search/niworkflowsthe multi-start searchniworkflows' arguments (15 0.1 0 10)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/principal-axesthe multi-start searchFrom the principal axesBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAffineInitializer/search/principal-axes-maskthe multi-start searchFrom the principal axes, a fixed maskBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/missing-maskthe multi-start searchA mask that cannot be read is no maskBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/fraction-above-onethe multi-start searchA radian fraction of 5 (clamped to 1)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/fraction-below-minimumthe multi-start searchA radian fraction of 0.001 (clamped to 0.01)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/local-search-textthe multi-start searchA local-search argument that is not read (ANTS-54; corrected: an error)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/local-search-zerothe multi-start searchA local search of 0 iterations (ANTS-54; corrected: start points only evaluated)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/local-search-negativethe multi-start searchA negative local search (ANTS-54; corrected: an error)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/local-search-20the multi-start searchA local search of 20 iterations, ANTs' own (only ITK-21 can differ)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/principal-axes-2the multi-start searchA principal-axes flag of 2 (true)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/dimension-text-prefixthe multi-start searchA dimension of 3x (std::stoi reads 3)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/2dthe multi-start search2D from the principal axesBit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAffineInitializer/search/2d-defaultsthe multi-start search2D, every defaultBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/4dthe multi-start search4D: the initial alignment onlyBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/4d-principal-axesthe multi-start search4D from the principal axesBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/obliquethe multi-start searchOblique grids, from the principal axesBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/txtthe multi-start searchA text transform fileBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/h5the multi-start searchAn HDF5 transform fileBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/search/extra-argumentsthe multi-start searchArguments after the mask are ignoredBit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/real/mni-2mmreal imagesniworkflows' arguments on the 2 mm MNI template and fMRIPrep's preprocessed T1w (ds000005)Bit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAffineInitializer/panel/fmriprep/ds000005-sub-01/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds000005-sub-01 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file; the replica − ANTs at 1 thread: 0 in a parameter

    7/7
    antsAffineInitializer/panel/fmriprep/ds000228-sub-pixar008/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds000228-sub-pixar008 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds000256-sub-CTS203/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds000256-sub-CTS203 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds001600-sub-1/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds001600-sub-1 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds002422-sub-02/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds002422-sub-02 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds002603-sub-01/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds002603-sub-01 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds003592-sub-280/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds003592-sub-280 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds004173-sub-603309/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds004173-sub-603309 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds004884-sub-M2079/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds004884-sub-M2079 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds004958-sub-fsm64ps/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds004958-sub-fsm64ps (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds005003-sub-RS001/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds005003-sub-RS001 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds005040-sub-002/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds005040-sub-002 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds005454-sub-16/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds005454-sub-16 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6
    antsAffineInitializer/panel/fmriprep/ds005752-sub-ON01016/anat_fit/register_template[MNI152NLin2009cAsym]/registration/antsAffineInitializerdata panel: fMRIPrep's callssMRIPrep's antsAffineInitializer inside the template registration node `anat_fit/register_template[MNI152NLin2009cAsym]/registration` of ds005752-sub-ON01016 (nipype's AffineInitializer arguments; the masked template and T1w made again exactly from the node's inputs)Bit-identical

    exit status, standard output, standard error, every file

    6/6

    Speed

    The replica of antsAI takes about half of ANTs’ time at one thread (17.5 s against 31.2 s on a 128 × 128 × 96 head, 79 s against 156 s on the 7 T MP2RAGE and the MNI template) and a sixth or less at eight threads, with ANTs’ memory. The replica of antsAffineInitializer takes ANTs’ time at one thread in reproduce (its check of ITK-21 searches twice), half of it in fix, and a quarter to a tenth at eight threads. Every transform is ANTs’ byte for byte: the benchmark (median of three runs; informal timings on a shared machine).

    Random testing

    Random testing runs random valid inputs and option combinations through the original program 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 comparison case, registered as a defect of the original, or reported. The latest long run:

    Random testing, the replica against the original: ANTs antsAI and antsAffineInitializer 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
    antsAffineInitializerreplica80706807·······
    antsAIreplica1,1020171,089··75··1
    • fixed in LARMORX antsAI (replica), 3 case(s): ANTs with CORE-MATH: standard error: antsAI’s start points that throw: each start point’s scales warnings come before its multi-start warning in ITK; LARMORX printed the multi-start warnings first. The problem now hands its warnings to the multi-start loop as they arise
    • fixed in LARMORX antsAI (replica), 2 case(s): ANTs with CORE-MATH: t.txt: a parameter of antsAI’s output transform at an exact tie between two shortest decimals: ITK’s double-conversion writes the even one, Rust’s shortest formatting rounded up; LARMORX’s ITK number writer now takes the even one (larmorx_io::itk_transform::itk_number)

    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 expected divergences: where ANTs never finishes or crashes (an empty blob list, ANTS-115; -b at more than one thread, ANTS-60), the replica refuses or gives ANTs’ one-thread result; where ANTs seeds its random sampling from the clock (ANTS-58), the replica is repeatable. The blob alignment’s corrections (ANTS-110, -111, -112, -114, -116 and ITK-60) are the replica’s default; upstream_defects="reproduce" gives ANTs’ result bit for bit.

    History

    • 2026-10-23: ANTS-116 decided by the project’s rules (the input makes sense, the result is wrong): corrected by default.
    • 2026-10-24 and 2026-10-28: antsAI’s full and standard comparisons made (their Generated lines); antsAffineInitializer’s full comparison on 2026-10-17 and its standard on 2026-10-28.
    • 2026-10-26: the data panel’s records of both programs.

    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.