Repository navigation
Home
1. MeshNet Model
-
A specialized deep learning architecture for volumetric brain MRI segmentation.
-
Flexible configuration for input channels, output classes, and training hyperparameters.
2. Federated Learning
-
NVFlare ensures privacy-preserving training across decentralized data sites.
-
Custom gradient aggregation methods optimize distributed model updates.
3. Data Pipeline
-
MRI data is stored in an SQLite database (mindboggle.db) with zlib compression.
-
Automatic dataset partitioning into training, validation, and testing subsets using splits.json.