MR-net - Distributed deep learning model for Medical Injuries classification.

Implementation of Mrnet using Keras with tensorflow backend in Jupyter notebook. It contains code for visuaization, model training and conversion of DICOM to numpy files (.npy).

Dataset used - MRNet

MRNet dataset is published by Stanford university for MRNet competition. It contains 1370 knee MRI's out of which 1,104 (80.6%) are abnormal exams, with 319 (23.3%) ACL tears and 508 (37.1%) meniscal tears. You can get dataset from Stanford Universty here.

Instructions

Change directory paths accordingly to run. Rest follow markdown in jupyter notebook.