(Registration comparision)Volumetric_quantification
In this notebook, we start at raw MRI scans in native space, and then using ANTs [1] (antsRegistrationSyNQuick.sh) to transform them to SIGMA space [2]. The brain structure volumes are computed using thoes files as well as SIGMA atlas.
Reference:
[1] Avants, Brian B., Nick Tustison, and Gang Song. “Advanced normalization tools (ANTS).” Insight j 2.365 (2009): 1-35.
[2] Barrière, D. A., et al. “The SIGMA rat brain templates and atlases for multimodal MRI data analysis and visualization.” Nature communications 10.1 (2019): 1-13.
** NOTE: **
Registration step takes several hours per MRI volume/scan. To avoid long waiting, we recommend run ANTs on high-performance PC/server, rather than on Colab.
We recommend users to view/check this notebook, which provides ANTs transformed MRI scans for volumetric quantification in longitudinal studies.
# Longitudinal Study Overview Adolescent rats (3, 6, 9, and 12 weeks old)
** Conclusion: **
Using *BEN* as the preprocessing pipeline before registration can observe the development of brain structure volumes, while others fail.
# 1.0 Install ANTs in Colab ANTs installation guide: https://www.suyogjadhav.com/misc/2019/03/28/Using-ANTs-package-on-Google-Colaboratory/
Download and unzip ANTs file.
[ ]:
!gdown 1N1Qx-R5tLCX5EhXPoPdyg6YvEkDtf-cD
Downloading...
From: https://drive.google.com/uc?id=1N1Qx-R5tLCX5EhXPoPdyg6YvEkDtf-cD
To: /content/ANTs-28-03-2019.7z
100% 316M/316M [00:02<00:00, 155MB/s]
[ ]:
!7z x ANTs-28-03-2019.7z
7-Zip [64] 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,2 CPUs Intel(R) Xeon(R) CPU @ 2.20GHz (406F0),ASM,AES-NI)
Scanning the drive for archives:
0M Scan 1 file, 315597527 bytes (301 MiB)
Extracting archive: ANTs-28-03-2019.7z
--
Path = ANTs-28-03-2019.7z
Type = 7z
Physical Size = 315597527
Headers Size = 2380
Method = LZMA2:24 BCJ
Solid = +
Blocks = 1
0% 0% 1 - bin/ANTS 1% 1 - bin/ANTS 1% 2 - bin/ANTSIntegrateVectorField 2% 2 - bin/ANTSIntegrateVectorField 2% 3 - bin/ANTSIntegrateVelocityField 3% 3 - bin/ANTSIntegrateVelocityField 3% 4 - bin/ANTSJacobian 4% 4 - bin/ANTSJacobian 4% 5 - bin/ANTSUseDeformationFieldToGetAffineTransform 5% 5 - bin/ANTSUseDeformationFieldToGetAffineTransform 5% 6 - bin/ANTSUseLandmarkImagesToGetAffineTransform 6% 6 - bin/ANTSUseLandmarkImagesToGetAffineTransform 6% 7 - bin/ANTSUseLandmarkImagesToGetBSplineDisplacementField 7% 7 - bin/ANTSUseLandmarkImagesToGetBSplineDisplacementField 7% 8 - bin/Atropos 8% 8 - bin/Atropos 8% 9 - bin/AverageAffineTransform 9% 9 - bin/AverageAffineTransform 9% 10 - bin/AverageAffineTransformNoRigid 10% 10 - bin/AverageAffineTransformNoRigid 11% 10 - bin/AverageAffineTransformNoRigid 11% 11 - bin/AverageImages 12% 11 - bin/AverageImages 12% 12 - bin/AverageTensorImages 12% 13 - bin/ClusterImageStatistics 13% 13 - bin/ClusterImageStatistics 14% 14 - bin/ComposeMultiTransform 14% 15 - bin/CompositeTransformUtil 15% 15 - bin/CompositeTransformUtil 16% 15 - bin/CompositeTransformUtil 16% 16 - bin/ConvertImage 17% 16 - bin/ConvertImage 17% 17 - bin/ConvertImagePixelType 18% 17 - bin/ConvertImagePixelType 18% 18 - bin/ConvertInputImagePixelTypeToFloat 19% 20 - bin/ConvertToJpg 20% 21 - bin/ConvertTransformFile 21% 21 - bin/ConvertTransformFile 21% 22 - bin/CopyImageHeaderInformation 22% 22 - bin/CopyImageHeaderInformation 22% 23 - bin/CreateDTICohort 23% 24 - bin/CreateDisplacementField 24% 24 - bin/CreateDisplacementField 25% 25 - bin/CreateImage 25% 27 - bin/CreateTiledMosaic 26% 28 - bin/CreateWarpedGridImage 27% 29 - bin/DeNrrd 28% 30 - bin/DenoiseImage 29% 31 - bin/ExtractRegionFromImage 30% 31 - bin/ExtractRegionFromImage 30% 33 - bin/ExtractSliceFromImage 31% 34 - bin/FitBSplineToPoints 32% 35 - bin/GetConnectedComponentsFeatureImages 33% 35 - bin/GetConnectedComponentsFeatureImages 33% 38 - bin/ImageCompare 34% 39 - bin/ImageIntensityStatistics 35% 40 - bin/ImageMath 36% 40 - bin/ImageMath 36% 41 - bin/ImageSetStatistics 37% 41 - bin/ImageSetStatistics 37% 42 - bin/KellyKapowski 38% 42 - bin/KellyKapowski 39% 43 - bin/KellySlater 39% 44 - bin/LabelClustersUniquely 40% 44 - bin/LabelClustersUniquely 40% 45 - bin/LabelGeometryMeasures 41% 45 - bin/LabelGeometryMeasures 41% 46 - bin/LabelOverlapMeasures 42% 46 - bin/LabelOverlapMeasures 42% 47 - bin/LaplacianThickness 43% 48 - bin/LesionFilling 44% 49 - bin/MeasureImageSimilarity 45% 49 - bin/MeasureImageSimilarity 45% 50 - bin/MeasureMinMaxMean 46% 50 - bin/MeasureMinMaxMean 46% 51 - bin/MemoryTest 47% 51 - bin/MemoryTest 47% 52 - bin/MultiplyImages 48% 53 - bin/N3BiasFieldCorrection 49% 53 - bin/N3BiasFieldCorrection 49% 54 - bin/N4BiasFieldCorrection 50% 54 - bin/N4BiasFieldCorrection 50% 55 - bin/NonLocalSuperResolution 51% 56 - bin/PasteImageIntoImage 51% 57 - bin/PermuteFlipImageOrientationAxes 52% 58 - bin/PrintHeader 53% 59 - bin/RebaseTensorImage 54% 59 - bin/RebaseTensorImage 54% 60 - bin/ReorientTensorImage 55% 60 - bin/ReorientTensorImage 55% 61 - bin/ResampleImage 56% 61 - bin/ResampleImage 57% 61 - bin/ResampleImage 57% 62 - bin/ResampleImageBySpacing 58% 63 - bin/ResetDirection 58% 64 - bin/SetDirectionByMatrix 59% 65 - bin/SetOrigin 60% 66 - bin/SetSpacing 61% 67 - bin/SimulateDisplacementField 62% 68 - bin/SmoothDisplacementField 63% 69 - bin/SmoothImage 64% 70 - bin/StackSlices 65% 72 - bin/SurfaceBasedSmoothing 66% 73 - bin/SurfaceCurvature 67% 74 - bin/TextureCooccurrenceFeatures 67% 75 - bin/TextureRunLengthFeatures 68% 76 - bin/ThresholdImage 69% 77 - bin/TileImages 70% 78 - bin/TimeSCCAN 71% 79 - bin/WarpImageMultiTransform 72% 79 - bin/WarpImageMultiTransform 72% 80 - bin/WarpTensorImageMultiTransform 73% 80 - bin/WarpTensorImageMultiTransform 73% 81 - bin/WarpTimeSeriesImageMultiTransform 74% 81 - bin/WarpTimeSeriesImageMultiTransform 74% 82 - bin/antsAI 75% 82 - bin/antsAI 76% 82 - bin/antsAI 76% 83 - bin/antsAffineInitializer 77% 83 - bin/antsAffineInitializer 77% 84 - bin/antsAlignOrigin 78% 84 - bin/antsAlignOrigin 78% 85 - bin/antsApplyTransforms 79% 85 - bin/antsApplyTransforms 80% 85 - bin/antsApplyTransforms 80% 86 - bin/antsApplyTransformsToPoints 81% 86 - bin/antsApplyTransformsToPoints 81% 87 - bin/antsJointFusion 82% 87 - bin/antsJointFusion 82% 88 - bin/antsJointTensorFusion 83% 89 - bin/antsLandmarkBasedTransformInitializer 84% 89 - bin/antsLandmarkBasedTransformInitializer 84% 90 - bin/antsMotionCorr 85% 90 - bin/antsMotionCorr 85% 91 - bin/antsMotionCorrDiffusionDirection 86% 91 - bin/antsMotionCorrDiffusionDirection 87% 91 - bin/antsMotionCorrDiffusionDirection 87% 92 - bin/antsMotionCorrStats 88% 92 - bin/antsMotionCorrStats 88% 93 - bin/antsRegistration 89% 93 - bin/antsRegistration 90% 93 - bin/antsRegistration 91% 93 - bin/antsRegistration 91% 94 - bin/antsSliceRegularizedRegistration 92% 94 - bin/antsSliceRegularizedRegistration 92% 95 - bin/antsTransformInfo 93% 95 - bin/antsTransformInfo 93% 96 - bin/antsUtilitiesTesting 94% 96 - bin/antsUtilitiesTesting 94% 97 - bin/compareTwoTransforms 95% 97 - bin/compareTwoTransforms 95% 98 - bin/iMath 96% 98 - bin/iMath 96% 99 - bin/itkTestDriver 97% 99 - bin/itkTestDriver 98% 100 - bin/sccan 98% 101 - bin/simpleSynRegistration 99% 101 - bin/simpleSynRegistration 100% 102 Everything is Ok
Folders: 1
Files: 101
Size: 1961813896
Compressed: 315597527
Now copy the entire contents of the newly created bin folder to /usr/local/bin/
[ ]:
!cp bin/* /usr/local/bin
Test the installation by running:
(Note: this ANTs version on Colab doesn’t provide antsRegistrationSyNQuick.sh.)
[ ]:
!which antsRegistration
!antsRegistration --version
/usr/local/bin/antsRegistration
ANTs Version: 3.0.0.0.dev35-g68094
Compiled: Mar 28 2019 10:37:32
# 2.0 Download raw MRI data and BEN
Here we use one rat with the first name_id [1]. Each rat was scaned 4 times (adolescent rats (3, 6, 9, and 12 weeks old)).
[1] name_id:
21d_cas_27_T2.nii.gz,
42d_cas_27_T2.nii.gz,
63d_cas_27_T2.nii.gz,
90d_cas_27_T2.nii.gz.
[ ]:
!git clone https://github.com/yu02019/BEN.git
Cloning into 'BEN'...
remote: Enumerating objects: 170, done.
remote: Counting objects: 100% (27/27), done.
remote: Compressing objects: 100% (25/25), done.
remote: Total 170 (delta 6), reused 7 (delta 0), pack-reused 143
Receiving objects: 100% (170/170), 6.48 MiB | 30.84 MiB/s, done.
Resolving deltas: 100% (50/50), done.
[ ]:
!gdown --id 1_M4wC8wHLWI776Hdkp2ktoNAE-cD7RMI
!unzip data_for_full_ants.zip
/usr/local/lib/python3.7/dist-packages/gdown/cli.py:131: FutureWarning: Option `--id` was deprecated in version 4.3.1 and will be removed in 5.0. You don't need to pass it anymore to use a file ID.
category=FutureWarning,
Downloading...
From: https://drive.google.com/uc?id=1_M4wC8wHLWI776Hdkp2ktoNAE-cD7RMI
To: /content/data_for_full_ants.zip
100% 630M/630M [00:04<00:00, 134MB/s]
Archive: data_for_full_ants.zip
creating: ANTs-sigma-afni2brain/
inflating: ANTs-sigma-afni2brain/21d_cas_27_T2.nii.gz0GenericAffine.mat
inflating: ANTs-sigma-afni2brain/21d_cas_27_T2.nii.gz1InverseWarp.nii.gz
inflating: ANTs-sigma-afni2brain/42d_cas_27_T2.nii.gz0GenericAffine.mat
inflating: ANTs-sigma-afni2brain/42d_cas_27_T2.nii.gz1InverseWarp.nii.gz
inflating: ANTs-sigma-afni2brain/63d_cas27_T2.nii.gz0GenericAffine.mat
inflating: ANTs-sigma-afni2brain/63d_cas27_T2.nii.gz1InverseWarp.nii.gz
inflating: ANTs-sigma-afni2brain/90d_cas27_T2.nii.gz0GenericAffine.mat
inflating: ANTs-sigma-afni2brain/90d_cas27_T2.nii.gz1InverseWarp.nii.gz
creating: ANTs-sigma-brain/
inflating: ANTs-sigma-brain/21d_cas_27_T2.nii.gz0GenericAffine.mat
inflating: ANTs-sigma-brain/21d_cas_27_T2.nii.gz1InverseWarp.nii.gz
inflating: ANTs-sigma-brain/42d_cas_27_T2.nii.gz0GenericAffine.mat
inflating: ANTs-sigma-brain/42d_cas_27_T2.nii.gz1InverseWarp.nii.gz
inflating: ANTs-sigma-brain/63d_cas_27_T2.nii.gz0GenericAffine.mat
inflating: ANTs-sigma-brain/63d_cas_27_T2.nii.gz1InverseWarp.nii.gz
inflating: ANTs-sigma-brain/90d_cas_27_T2.nii.gz0GenericAffine.mat
inflating: ANTs-sigma-brain/90d_cas_27_T2.nii.gz1InverseWarp.nii.gz
creating: Atlas-brain/
inflating: Atlas-brain/R2_SIGMA_Anatomical_Brain_Atlas.nii.gz
inflating: Atlas-brain/SIGMA_InVivo_Brain_Template_Masked.nii.gz
creating: atlas_woSS/
inflating: atlas_woSS/SIGMA_InVivo_Brain_Template.nii.gz
creating: pipeline-afni-src/
inflating: pipeline-afni-src/21d_cas_27_T2.nii.gz
inflating: pipeline-afni-src/42d_cas_27_T2.nii.gz
inflating: pipeline-afni-src/63d_cas_27_T2.nii.gz
inflating: pipeline-afni-src/90d_cas_27_T2.nii.gz
creating: pipeline-BEN-src/
inflating: pipeline-BEN-src/21d_cas_27_T2.nii.gz
inflating: pipeline-BEN-src/42d_cas_27_T2.nii.gz
inflating: pipeline-BEN-src/63d_cas_27_T2.nii.gz
inflating: pipeline-BEN-src/90d_cas_27_T2.nii.gz
creating: pipeline-woSkullStrip-src/
inflating: pipeline-woSkullStrip-src/21d_cas_27_T2.nii.gz
inflating: pipeline-woSkullStrip-src/42d_cas_27_T2.nii.gz
inflating: pipeline-woSkullStrip-src/63d_cas_27_T2.nii.gz
inflating: pipeline-woSkullStrip-src/90d_cas_27_T2.nii.gz
For scan ‘21d_cas_27_T2.nii.gz’, antsRegistration will generate output files look like this:
Filename |
Use |
|---|---|
{name_id}_0GenericAffine.mat |
** Linear transform matrix ** |
{name_id}_1InverseWarp.nii.gz |
** Inverse nonlinear trans mat ** |
{name_id}_1Warp.nii.gz |
Nonlinear trans mat |
{name_id}_InverseWarped.nii.gz |
Inverse warped/transformed MRI scan |
{name_id}_Warped.nii.gz |
Warped/transformed MRI scan |
As registration takes long time for all data and different method pipelines (>48 hours), we provide inverse transformation matrix here (red in Table), in order to transform atlas labels to native space.
These trans mat were acquired using following command. With these mat, we can also tranform label/mask/annotation/etc between two space bidirectionally. In other word, this is how to do atlas-based brain structure segmentation.
3.0 ANTs registration via BEN interface
Just using following commands via BEN’s interface:
# python BEN/BEN_infer.py -i input_folder -o output_folder # BEN skull stripping
python BEN/interface/python/utils_ants_os.py -i input_folder -o output_folder -ref fixed_image
python BEN/interface/python/utils_trans_os.py -i input_folder -o output_folder -ref fixed_image -mat trans_mat_folder
or use original ANTs
# 3dSkullStrip -input input_folder -orig_vol -prefix xxx -DAFNI_NIFTI_TYPE_WARN=YES # BEN skull stripping
antsRegistration -d 3 -o 'test.nii.gz' -f 'Atlas-brain/SIGMA_InVivo_Brain_Template_Masked.nii.gz' -m 'pipeline-BEN-src/21d_cas_27_T2.nii.gz'
antsApplyTransforms -d 3 -i 'Atlas-brain/R2_SIGMA_Anatomical_Brain_Atlas.nii.gz' -r 'pipeline-BEN-src/21d_cas_27_T2.nii.gz' -n NearestNeighbor -t ['21d_cas_27_T2.nii.gz0GenericAffine.mat', 1] -t '21d_cas_27_T2.nii.gz1InverseWarp.nii.gz' -o 'output_trans.nii.gz'
[ ]:
ls
ANTs-28-03-2019.7z atlas_woSS/ pipeline-afni-src/
ANTs-sigma-afni2brain/ BEN/ pipeline-BEN-src/
ANTs-sigma-brain/ bin/ pipeline-woSkullStrip-src/
Atlas-brain/ data_for_full_ants.zip sample_data/
3.1 BEN pipeline transform
[ ]:
!python BEN/interface/python/utils_trans_os.py -i 'pipeline-BEN-src/*' -o 'Native-label-brain/' -ref 'Atlas-brain/R2_SIGMA_Anatomical_Brain_Atlas.nii.gz' -mat 'ANTs-sigma-brain/'
Creating new folder: Native-label-brain/
0% 0/4 [00:00<?, ?it/s]21d_cas_27_T2.nii.gz
25% 1/4 [00:01<00:05, 1.67s/it]42d_cas_27_T2.nii.gz
50% 2/4 [00:03<00:03, 1.64s/it]63d_cas_27_T2.nii.gz
75% 3/4 [00:04<00:01, 1.62s/it]90d_cas_27_T2.nii.gz
100% 4/4 [00:06<00:00, 1.61s/it]
Done!
total time: 7.46518874168396
avg time: 1.86629718542099
## 3.2 AFNI pipeline transform
[ ]:
!python BEN/interface/python/utils_trans_os.py -i 'pipeline-afni-src/*' -o 'Native-label-afni-brain/' -ref 'Atlas-brain/R2_SIGMA_Anatomical_Brain_Atlas.nii.gz' -mat 'ANTs-sigma-afni2brain/'
Creating new folder: Native-label-afni-brain/
0% 0/4 [00:00<?, ?it/s]21d_cas_27_T2.nii.gz
25% 1/4 [00:01<00:04, 1.51s/it]42d_cas_27_T2.nii.gz
50% 2/4 [00:03<00:03, 1.52s/it]63d_cas_27_T2.nii.gz
lose trans mat!
90d_cas_27_T2.nii.gz
lose trans mat!
100% 4/4 [00:03<00:00, 1.31it/s]
Done!
total time: 4.067110538482666
avg time: 1.0167776346206665
3.3 Visualization
[ ]:
!pip install medpy
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting medpy
Downloading MedPy-0.4.0.tar.gz (151 kB)
|████████████████████████████████| 151 kB 5.2 MB/s
Requirement already satisfied: scipy>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from medpy) (1.7.3)
Requirement already satisfied: numpy>=1.11.0 in /usr/local/lib/python3.7/dist-packages (from medpy) (1.21.6)
Collecting SimpleITK>=1.1.0
Downloading SimpleITK-2.1.1.2-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (48.4 MB)
|████████████████████████████████| 48.4 MB 1.4 MB/s
Building wheels for collected packages: medpy
Building wheel for medpy (setup.py) ... done
Created wheel for medpy: filename=MedPy-0.4.0-cp37-cp37m-linux_x86_64.whl size=754449 sha256=d2a1ec6af92ca1cc5cb7f936996aeea96a7b7ced7e7b5c906d53431c59d90a3e
Stored in directory: /root/.cache/pip/wheels/b0/57/3a/da1183f22a6afb42e11138daa6a759de233fd977a984333602
Successfully built medpy
Installing collected packages: SimpleITK, medpy
Successfully installed SimpleITK-2.1.1.2 medpy-0.4.0
[ ]:
cd BEN
/content/BEN
[ ]:
import matplotlib.pyplot as plt
from utils.transparent_imshow import transp_imshow
from utils.visualization import load_slice
def plot_segmentation(raw, baseline, BEN, task, hspace=-0.7):
fig = plt.figure(
figsize=(16,9)
)
timepoint_count = raw.shape[0]
for i in range(timepoint_count):
plt.subplot(3, timepoint_count, i + 1)
plt.imshow(raw[i], cmap='gray')
plt.xticks([]), plt.yticks([])
for i in range(timepoint_count):
plt.subplot(3, timepoint_count, timepoint_count + i + 1)
plt.imshow(raw[i], cmap='gray')
transp_imshow(baseline[i], cmap='jet', alpha=0.6)
plt.xticks([]), plt.yticks([])
for i in range(timepoint_count):
plt.subplot(3, timepoint_count, timepoint_count * 2 + i + 1)
plt.imshow(raw[i], cmap='gray')
transp_imshow(BEN[i], cmap='jet', alpha=0.6)
plt.xticks([]), plt.yticks([])
plt.subplots_adjust(hspace=hspace, wspace=0.0)
fig.suptitle('Plot of different methods on cross-{} task'.format(task), fontsize=14)
plt.show()
[ ]:
raw, baseline, BEN = load_slice(r'../pipeline-BEN-src',
r'../Native-label-afni-brain/',
r'../Native-label-brain/',
scans_num=1)
plot_segmentation(raw, baseline, BEN, task='pipeline method', hspace=-0.7)
/content/BEN/utils/transparent_imshow.py:50: RuntimeWarning: invalid value encountered in true_divide
color = (data-vmin)/(vmax-vmin)
/content/BEN/utils/transparent_imshow.py:53: RuntimeWarning: invalid value encountered in true_divide
transparency = tmax*(data-tvmin)/(tvmax-tvmin)
Volumetric_quantification_26_1.png)
Registration results:
From top raw to the third row: Raw image, AFNI pipeline, BEN pipeline.
The red circles circles indicate some representative brain regions after atlas registration. It can be seen that the results using BEN match the atlas, but using AFNI produce some errors.
# 4.0 Conclusion BEN performs much better than using AFNI for the atlas registration, thus benifiting volumetric quantifications.
[ ]: