DeepMSMT-CSD
Learning multi-shell MT-CSD FODs from single-shell DW-MRI
This repository contains the code written for learning multi-tissue constrained spherical deconvolution (MT-CSD) multi-shell fiber orientation distribution (FOD) reconstruction from single-shell DW-MRI, released with the 2020 SPIE paper. It includes the model training and inference entry points (main_files/), data generation utilities (data_gens/), and model weights (models/).
Usage
The workflow is Python-based: clean and generate training data with the scripts in data_cleaning_stuff/ and data_gens/, train with main_files/main_patch_smt_dl.py (or the volume-fraction variant main_patch_volfrac_dl.py), and evaluate with main_files/main_scsd_acc.py. Model weights are in models/. See the repository for environment setup; no pre-built container is currently published, so run the code from a standard deep-learning Python environment.
Vishwesh Nath, Sudhir K. Pathak, Kurt G. Schilling, Walter Schneider, and Bennett A. Landman. “Deep learning estimation of multi-tissue constrained spherical deconvolution with limited single shell DW-MRI” In Medical Imaging 2020: Image Processing, International Society for Optics and Photonics, 2020.