MRI Brain Scan Classification Guide | FastSurfer, XGBoost, 3D CNNs
summary
Step-by-step pipeline shows extracting volumetric features with FastSurfer + XGBoost pipeline (FastSurfer segments a brain in ~1 minute versus ~10 hours for FreeSurfer), preprocessing with skull‑stripping, registration to MNI152 and histogram intensity normalization, then training XGBoost for sex and age. A separate approach trains a pseudo-3D PyTorch CNN (pseudo‑3D blocks to save memory) on registered 3D scans with GPU; models converge and yield about 0.8 balanced accuracy for sex and ~7 years MAE for age, comparable between methods. Docker and GPU recommended.
AI-written summary · about
chapters (12)
- 0:00MRI brain scan classification
- 5:29MRI data processing theory
- 11:34Tabular data extraction
- 22:27FastSurfer preprocessing demo
- 26:44Combining tabular and demographic data
- 29:56XGBoost for sex and age
- 38:563D CNNs deep learning
- 53:10Preprocessing MRI scans for convolutional networks
- 1:00:03Pseudo 3D CNN training
- 1:11:23Model evaluation and results
- 1:14:04Future research directions
- 1:18:11Conclusion and Q&A
related talks
48mNeural Networks and Brains
15mBrain Tumor Detection and Classification Using Deep Learning Models
49mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
42mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
25mBig Brains, Small Targets: Whole-Brain Image Analysis with Python
27mCelebrating Neurodiversity Through Aesthetic Data Visualization
43mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
50mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics