7. Voxel Based Morphometry Workflow

7.1. Introduction

Voxel-Based Morphometry (VBM) is a widely used neuroimaging technique for assessing structural differences in brain anatomy across individuals or groups. It provides an automated, whole‑brain approach to quantifying regional gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) volumes from T1-weighted MRI scans. Unlike manual or region‑of‑interest methods, VBM does not require predefined anatomical boundaries, making it particularly well suited for exploratory analyses and large-scale studies.

7.2. Requirements

CPU	RAM
1	5 GB

7.3. Description

Processing Steps

• Initial Image Preparation CAT12 [1] applies several preparatory steps to improve image quality and prepare the data for segmentation: bias-field correction, affine registration to MNI space, and noise and intensity normalization to improve tissue contrast and reduce local intensity variations.

• Tissue Segmentation CAT12 [1] performs an advanced segmentation of the brain into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). The pipeline utilizes adaptive local segmentation (LAS) for enhanced boundary detection, graph-cut refinement to sharpen tissue borders, partial volume estimation to accurately model voxels containing mixed tissue types, white matter hyperintensity correction, and Markov Random Field smoothing to minimize isolated misclassifications.

• Spatial Normalization The segmented tissue maps are registered to an MNI-space template using a high-dimensional Geodesic Shooting algorithm. Both forward and inverse deformation fields are saved to enable seamless data transformation between native and template space.

• Modulation To preserve local tissue volumes following spatial normalization, tissue maps are adjusted using the Jacobian determinants of the deformation fields. This modulation step ensures that the final voxel intensities reflect regional tissue volume rather than tissue concentration.

• Resampling All normalized images are resampled to 1.5 mm isotropic resolution.

• ROI-Based Morphometry Regional volumetric and surface metrics are extracted using a comprehensive suite of atlases, including Neuromorphometrics, LPBA40, Hammers, AAL3, Julich Brain, COBRA, the Mori white‑matter atlas, and the Anatomy toolbox. For surface-based parcellations, such as the Schaefer 100/200/400/600 parcels, CAT12 extracts regional cortical thickness and surface area measures.

Longitudinal Processing Steps

• Intra‑subject realignment All time points for a participant are rigidly aligned to each other using an inverse-consistent registration fluid dynamics model to remove spatial differences caused by head position rather than true anatomical change.

• Creation of an unbiased within‑subject template CAT12 builds an unbiased subject‑specific anatomical template and a customized Tissue Probability Map (TPM) by averaging all time points equally. This template serves as a stable, session-independent reference for all subsequent processing.

• Bias correction and intensity normalization Each time point undergoes adaptive inhomogeneity correction and intensity normalization relative to the subject‑specific template, significantly reducing scanner noise and session‑to‑session signal variability.

• Longitudinal segmentation GM, WM, and CSF are segmented using priors derived from the subject‑specific template. This improves tissue classification consistency across time points.

• Spatial Normalization All time points are nonlinearly registered to the subject‑specific template, which is then mapped to the MNI standard space using a high-dimensional Geodesic Shooting algorithm. This two‑stage deformation strategy maximizes sensitivity to subtle, longitudinal structural changes.

• Modulation To preserve local tissue volumes after spatial normalization, CAT12 applies modulation to the GM, and WM maps using the Jacobian determinants of the deformation fields. This step ensures that the final voxel values reflect regional tissue volume rather than concentration.

• Resampling All normalized images are resampled to 1.5 mm isotropic resolution.

Quality Control:

• Correlation score For each image, we compute its correlation with the DARTEL template. Images are then sorted in ascending order of this score, allowing potential outliers to be easily identified.

• Manual inspection Following the correlation-based ranking, T1w images at the lower end of the distribution are manually reviewed in-house.

• Thresholding Both Noise Contrast Ratio (NCR) and Image Quality Rating (IQR) are thresholded at a maximum value of 4.5. Images with NCR > 4.5 or IQR > 4.5 are flagged as low‑quality.

7.4. Outputs

The vbm directory contains subject-level results, longitudinal results, group-level results, logs, and quality-control outputs. The structure is organized following the brainprep ontology.

vbm/
├── dataset_description.json
├── figures
│   ├── histogram_IQR.png
│   ├── histogram_mean_correlation.png
│   └── histogram_NCR.png
├── log
│   └── report_<timestamp>.rst
├── longitudinal
│   ├── figures
│   │   ├── histogram_IQR.png
│   │   ├── histogram_mean_correlation.png
│   │   └── histogram_NCR.png
│   ├── morphometry
│   │   ├── <atlas_name>_cat12_vbm_roi.tsv
│   │   └── cat12_vbm_total_volumes.tsv
│   ├── quality_check
│   │   ├── group_stats.tsv
│   │   └── mean_correlations.tsv
│   └── subjects
│       └── sub-<sub_id>
│           ├── cat12vbm_matlabbatch.m
│           ├── log
│           ├── ses-01
│           │   ├── avg_sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   ├── label
│           │   │   ├── catROI_rsub-<sub_id>_ses-01_<bids_keys>_T1w.mat
│           │   │   └── catROI_rsub-<sub_id>_ses-01_<bids_keys>_T1w.xml
│           │   ├── mri
│           │   │   ├── avg_y_rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── mean_mwp1rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── mean_mwp2rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── mwp1rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── mwp2rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── p0avg_sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── p0rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── p1rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── p2rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── rp1avg_sub-<sub_id>_ses-01_<bids_keys>_T1w_affine.nii
│           │   │   ├── rp2avg_sub-<sub_id>_ses-01_<bids_keys>_T1w_affine.nii
│           │   │   ├── rp3avg_sub-<sub_id>_ses-01_<bids_keys>_T1w_affine.nii
│           │   │   ├── tpidiff_mwp1rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── tpidiff_mwp2rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── tprdiff_mwp1rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   ├── tprdiff_mwp2rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   │   └── y_rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   ├── report
│           │   │   ├── cat_avg_sub-<sub_id>_ses-01_<bids_keys>_T1w.mat
│           │   │   ├── cat_avg_sub-<sub_id>_ses-01_<bids_keys>_T1w.xml
│           │   │   ├── catlog_avg_sub-<sub_id>_ses-01_<bids_keys>_T1w.txt
│           │   │   ├── catlog_rsub-<sub_id>_ses-01_<bids_keys>_T1w.txt
│           │   │   ├── catlongreportj_mwp1rsub-<sub_id>_ses-01_<bids_keys>_T1w.jpg
│           │   │   ├── catlongreport_mwp1rsub-<sub_id>_ses-01_<bids_keys>_T1w.pdf
│           │   │   ├── catlong_sub-<sub_id>_ses-01_<bids_keys>_T1w.mat
│           │   │   ├── catlong_sub-<sub_id>_ses-01_<bids_keys>_T1w.xml
│           │   │   ├── catreport_avg_sub-<sub_id>_ses-01_<bids_keys>_T1w.pdf
│           │   │   ├── catreportj_avg_sub-<sub_id>_ses-01_<bids_keys>_T1w.jpg
│           │   │   ├── catreportj_rsub-<sub_id>_ses-01_<bids_keys>_T1w.jpg
│           │   │   ├── catreport_rsub-<sub_id>_ses-01_<bids_keys>_T1w.pdf
│           │   │   ├── cat_rsub-<sub_id>_ses-01_<bids_keys>_T1w.mat
│           │   │   └── cat_rsub-<sub_id>_ses-01_<bids_keys>_T1w.xml
│           │   ├── rsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   ├── sanlm_sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   ├── sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           │   └── wavg_sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
│           └── ses-02
│               ├── label
│               │   ├── catROI_rsub-<sub_id>_ses-02_<bids_keys>_T1w.mat
│               │   └── catROI_rsub-<sub_id>_ses-02_<bids_keys>_T1w.xml
│               ├── mri
│               │   ├── mwp1rsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
│               │   ├── mwp2rsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
│               │   ├── p0rsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
│               │   ├── p1rsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
│               │   ├── p2rsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
│               │   └── y_rsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
│               ├── report
│               │   ├── catlog_rsub-<sub_id>_ses-02_<bids_keys>_T1w.txt
│               │   ├── catreportj_rsub-<sub_id>_ses-02_<bids_keys>_T1w.jpg
│               │   ├── catreport_rsub-<sub_id>_ses-02_<bids_keys>_T1w.pdf
│               │   ├── cat_rsub-<sub_id>_ses-02_<bids_keys>_T1w.mat
│               │   └── cat_rsub-<sub_id>_ses-02_<bids_keys>_T1w.xml
│               ├── rsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
│               ├── sanlm_sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
│               └── sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
├── morphometry
│   ├── <atlas_name>_cat12_vbm_roi.tsv
│   └── cat12_vbm_total_volumes.tsv
├── quality_check
│   ├── group_stats.tsv
│   └── mean_correlations.tsv
└── subjects
    └── sub-<sub_id>
        ├── ses-01
        │   ├── cat12vbm_matlabbatch_run-<run_id>.m
        │   ├── label
        │   │   ├── catROI_sub-<sub_id>_ses-01_<bids_keys>_T1w.mat
        │   │   └── catROI_sub-<sub_id>_ses-01_<bids_keys>_T1w.xml
        │   ├── log
        │   ├── mri
        │   │   ├── iy_sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   ├── mwp1sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   ├── mwp2sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   ├── mwp3sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   ├── p0sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   ├── p1sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   ├── p2sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   ├── p3sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   ├── wmsub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   │   └── y_sub-<sub_id>_ses-01_<bids_keys>_T1w.nii
        │   └── report
        │       ├── catlog_sub-<sub_id>_ses-01_<bids_keys>_T1w.txt
        │       ├── catreportj_sub-<sub_id>_ses-01_<bids_keys>_T1w.jpg
        │       ├── catreport_sub-<sub_id>_ses-01_<bids_keys>_T1w.pdf
        │       ├── cat_sub-<sub_id>_ses-01_<bids_keys>_T1w.mat
        │       └── cat_sub-<sub_id>_ses-01_<bids_keys>_T1w.xml
        └── ses-02
            ├── cat12vbm_matlabbatch_run-<run_id>.m
            ├── label
            │   ├── catROI_sub-<sub_id>_ses-02_<bids_keys>_T1w.mat
            │   └── catROI_sub-<sub_id>_ses-02_<bids_keys>_T1w.xml
            ├── log
            ├── mri
            │   ├── iy_sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── mwp1sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── mwp2sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── mwp3sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── nsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── p0sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── p1sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── p2sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── p3sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   ├── wmsub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            │   └── y_sub-<sub_id>_ses-02_<bids_keys>_T1w.nii
            └── report
                ├── catlog_sub-<sub_id>_ses-02_<bids_keys>_T1w.txt
                ├── catreportj_sub-<sub_id>_ses-02_<bids_keys>_T1w.jpg
                ├── catreport_sub-<sub_id>_ses-02_<bids_keys>_T1w.pdf
                ├── cat_sub-<sub_id>_ses-02_<bids_keys>_T1w.mat
                └── cat_sub-<sub_id>_ses-02_<bids_keys>_T1w.xml

Description of contents:

• dataset_description.json Metadata describing the process, including versioning and processing information.

• figures Contains quality measure histograms extracted by the group-level-vbm command.

• morphometry/<atlas_name>_cat12_vbm_roi.tsv Tissue volumes per <atlas_name>’s ROI. We extract the information for all atlases made available by cat12, whose list can be found in the cat12 documentation.

• morphometry/cat12_vbm_total_volumes.tsv Total intracranial volume (TIV) and total tissue volumes per subject.

• quality_check/group_stats.tsv Quality metrics extracted by the group-level-vbm pipeline. See its doc for more info.

• quality_check/mean_correlations.tsv Mean correlation of each image to the used template. Should be close to 1 if alignment went right.

• log/report_<timestamp>.rst Contains group-level workflow steps and parameters.

• subjects/sub-<id>/ses-<id> subject-level-vbm pipeline outputs, consisting in cat12 outputs. If multiple images for a single session, the outputs are in the same folder.

  • cat12vbm_matlabbatch_run-<run_id>.m Parameters file used by cat12. If no run_id for the image, it is generated using a hash algorithm on the input filename.

  • label Files containing the image ROI volumes. They are summarized for all subjects in ‘morphometry’ folder.

  • log Cat12 logs.

  • mri Cat12 output images. The naming convention is described in the cat12 documentation.

  • report Summary reports generated by cat12. They contain process parameters, metrics and quality metrics.

• longitudinal Same organisation as the main output folder, but for the longitudinal-vbm pipeline outputs.

7.5. Featured examples

VBM

Explore how to perform this analysis.

../auto_examples/workflows/plot_vbm.html

7.6. References

[1] (1,2)

Christian Gaser, Robert Dahnke, Paul M Thompson, Florian Kurth, Eileen Luders, and The Alzheimer’s Disease Neuroimaging Initiative. CAT: a computational anatomy toolbox for the analysis of structural MRI data. Gigascience, January 2024.