We introduce an open-source tool for automated subregional assessment of knee cartilage degradation using quantitative T2 relaxometry and deep learning.
Manual or semi-automated segmentation of cartilage in quantitative MRI scans is a necessary step in assessing early changes in cartilage health. The aim of this work was to develop a fully automated femoral cartilage segmentation model and to evaluate the model's ability to measure subregional T2 values longitudinally.
When using the tool, please cite the following paper:
Thomas, Kevin A., Dominik Krzemiński, Łukasz Kidziński, Rohan Paul, Elka B. Rubin, Eni Halilaj, Marianne S. Black, Akshay Chaudhari, Garry E. Gold, and Scott L. Delp. "Open source software for automatic subregional assessment of knee cartilage degradation using quantitative T2 relaxometry and deep learning." Cartilage 13, no. 1_suppl (2021): 747S-756S.
Link to paper: https://journals.sagepub.com/doi/abs/10.1177/19476035211042406?journalCode=cara
Link to code: https://github.com/kathoma/AutomaticKneeMRISegmentation
This software provides the following automated functionality for multi-echo spin echo T2-weighted knee MRIs:
Segmentation of femoral cartilage
Projection of the femoral cartilage onto a 2D plane
Division of the projected cartilage into 12 subregions along medial-lateral, superficial-deep, and anterior-central-posterior boundaries
Calculation of the average T2 value in each subregion
Calculation of the change in average T2 value over time for each subregion (if 2 imaging time points are available for a given person)
Comparison of results across different readers/models
FullPipeline.ipynb walks through an example of how to use the full pipeline to analyze individual images, calculate changes in a patient over time, and compare results for segmentations from different readers.
Requires CUDA Version 9.0.176. Tested with CUDA 9.0 and cudnn 7.3.0 in Ubuntu 18.04.