How mahdisabetkish/brainssl works
brainssl is a research toolkit for label-efficient brain MRI and resting-state fMRI modeling. Users prepare public imaging data, run either the structural-MRI pretraining/fine-tuning or connectome-classification study, then evaluate results and optionally audit or export models. The graph separates those domain paths and includes the reporting layer. Relationships are limited to README claims and visible import references in the supplied source index; unsampled implementations are represented by their exact source paths, but uncertain internal calls are omitted.
Main entry points: cli.py CLI router · study.py Connectome study · train.py Pretrain and tune · smri.py Trust audits
Components 22
Command entry
- CLI router
src/brainssl/cli.py - Data download
src/brainssl/data/download.py
Structural MRI
- Volume preparation
src/brainssl/smri/preprocess.py - Twin masking
src/brainssl/smri/masking.py - Bilateral U-Net
src/brainssl/smri/model.py - Pretrain and tune
src/brainssl/smri/train.py - Segmentation inference
src/brainssl/smri/infer.py - Segmentation evaluation
src/brainssl/smri/evaluate.py - Segmentation metrics
src/brainssl/smri/metrics.py
Functional MRI
- ABIDE time series
src/brainssl/fmri/abide.py - Connectome estimation
src/brainssl/fmri/connectivity.py - Scan windows
src/brainssl/fmri/windows.py - Classifiers
src/brainssl/fmri/methods.py - Fold evaluation
src/brainssl/fmri/evaluate.py - Connectome study
src/brainssl/fmri/study.py
Analysis and delivery
- Report commands
src/brainssl/report/cli.py - Result collection
src/brainssl/report/collect.py - Dashboard data
src/brainssl/report/dashboard.py - Trust audits
src/brainssl/trust/smri.py - Model export
src/brainssl/export/cli.py - Statistical tests
src/brainssl/stats.py
Other components
- Researcher
Connections 29
- Researcher → CLI router: runs commands
src/brainssl/cli.py - CLI router → Data download: invokes download
src/brainssl/cli.py - CLI router → Volume preparation: invokes preparation
src/brainssl/cli.py - CLI router → Pretrain and tune: starts training
src/brainssl/cli.py - Pretrain and tune → Twin masking: uses masking
src/brainssl/smri/train.py - Pretrain and tune → Bilateral U-Net: trains model
src/brainssl/smri/train.py - Pretrain and tune → Volume preparation: loads volumes
src/brainssl/smri/train.py - CLI router → Segmentation evaluation: invokes evaluation
src/brainssl/cli.py - Segmentation evaluation → Bilateral U-Net: evaluates model
src/brainssl/smri/evaluate.py - Segmentation evaluation → Segmentation metrics: scores predictions
- CLI router → ABIDE time series: starts fMRI workflow
src/brainssl/cli.py - ABIDE time series → Connectome estimation: estimates matrices
- ABIDE time series → Scan windows: supplies time series
src/brainssl/fmri/windows.py - Scan windows → Connectome estimation: estimates window views
src/brainssl/fmri/windows.py - CLI router → Connectome study: starts study
src/brainssl/cli.py - Connectome study → ABIDE time series: loads subject data
src/brainssl/fmri/study.py - Connectome study → Connectome estimation: uses embeddings
src/brainssl/fmri/study.py - Connectome study → Classifiers: compares classifiers
src/brainssl/fmri/study.py - Connectome study → Fold evaluation: uses folds
src/brainssl/fmri/study.py - CLI router → Report commands: routes reporting
src/brainssl/cli.py - Report commands → Result collection: collects results
src/brainssl/report/cli.py - Report commands → Dashboard data: builds dashboard data
src/brainssl/cli.py - Result collection → Statistical tests: computes statistics
src/brainssl/report/collect.py - CLI router → Trust audits: routes trust audit
src/brainssl/cli.py - CLI router → Model export: routes export
src/brainssl/cli.py - Model export → Bilateral U-Net: exports segmentation model
- Trust audits → Segmentation inference: audits predictions
src/brainssl/trust/smri.py - Trust audits → Volume preparation: loads test scans
src/brainssl/trust/smri.py - Trust audits → Statistical tests: tests comparisons
src/brainssl/trust/smri.py