How tranvanhuy-hichan/brain-mri-segmentation works
NeuroSeg is a research scaffold for brain-MRI tumor segmentation, with a dataset-to-model workflow and a separate native review application. The documented main path validates/imports NIfTI cases, then invokes nnU-Net planning, training and prediction; evaluation is planned. A second branch supports slice and 3D review of MRI volumes and masks. Kaggle files provide CPU preprocessing and budgeted pilot-training stages. Some pipeline and integration details are documented but not excerpted, so those links cite the README; the graph does not infer unsampled calls.
Main entry points: pipeline.sh Pipeline wrappers · app.py Qt workspace · view3d.py 3D viewer · prepare_cpu.py CPU preprocessing
Components 17
Dataset preparation
- NIfTI importer
scripts/prepare_dataset.py - nnU-Net dataset
Model workflow
- Pipeline wrappers
scripts/pipeline.sh - Planning and preprocessing
- Fold training
- Mask prediction
- Evaluation
Kaggle pilot
- CPU preprocessing
kaggle/prepare_cpu.py - Budgeted training
kaggle/train_budget.py - Kaggle launcher
kaggle/launch_after_prepare.py
Desktop review
- Qt workspace
desktop/app.py - NIfTI volume
desktop/volume.py - Slice rendering
desktop/volume.py - 3D viewer
desktop/view3d.py - MRI volume rendering
desktop/anatomy.py - Mask surfaces
desktop/surfaces.py
Other components
- Researcher
Connections 17
- Researcher → NIfTI importer: prepares cases
README.md - NIfTI importer → nnU-Net dataset: writes dataset
README.md - Researcher → Pipeline wrappers: runs stages
README.md - Pipeline wrappers → Planning and preprocessing: invokes planning
README.md - Pipeline wrappers → Fold training: invokes training
README.md - Pipeline wrappers → Mask prediction: invokes prediction
README.md - Mask prediction → Evaluation: planned evaluation
README.md - Researcher → CPU preprocessing: starts preparation
README.md - CPU preprocessing → Budgeted training: precedes pilot
README.md - Kaggle launcher → CPU preprocessing: checks completion
kaggle/launch_after_prepare.py - Kaggle launcher → Budgeted training: submits kernel
kaggle/launch_after_prepare.py - Researcher → Qt workspace: opens viewer
README.md - Qt workspace → NIfTI volume: loads volumes
desktop/app.py - Qt workspace → 3D viewer: hosts viewer
desktop/app.py - NIfTI volume → Slice rendering: renders slices
desktop/volume.py - 3D viewer → MRI volume rendering: builds anatomy
desktop/view3d.py - 3D viewer → Mask surfaces: builds surfaces
desktop/view3d.py