torchio

Why we included this project

TorchIO does the data preparation between raw medical scans and a PyTorch training loop. It reads common neuroimaging formats, applies intensity and spatial transforms as augmentation, and cuts large 3D volumes into patches so segmentation or classification networks can train on scans that would not fit in memory whole. The transforms are composable PyTorch-style operations, which keeps preprocessing inside the pipeline instead of scattered across standalone scripts. Teams working with MRI, CT, or other volumetric scans get a practical way to standardize data preparation and add realistic variation without hand-writing custom loaders, and the codebase doubles as a clear example of how a modern medical-imaging pipeline is put together.