Mitral Valve Segmentation using Robust Nonnegative Matrix Factorization

This repository is the official MATLAB implementation of Mitral Valve Segmentation using Robust Nonnegative Matrix Factorization.

Some results of the proposed methods are shown in: https://drive.google.com/drive/folders/1Scpy54x0-_zXpP3bXZIYJPitN2kMUyB1?usp=sharing

Get Started

• Please insert the videos that should be used for mitral valve segmentation into the folder: ./data/original/
• Create the ground truth data by running the file ./tools/createGroundTruth.m, which will be saved into ./data/ground_truth/ and ./data/ground_truth_window/
• Run ./main.m to start the algorithm
• Run ./tools/showResults.m to analyze the segmentation results

How to run main.m

Robust Nonnegative Matrix Factorization

• To run one of the Robust Nonnegative Matrix Factorization (RNMF) methods, please run ids = rnmf([methodname], [numberofvideos]);
• Instead of methodname, insert one of the following methods: 'robustNMF', 'robustBreg', 'robustNMF_excludeWHS' or 'robustNMF_excludeWHS_Breg'
• Replace numberofvideos by the number of videos you want to segment
• Please adapt the settings for RNMF in ./rnmf.m

Segmentation

• To apply the segmentation methods, please run segment('segmentCVPlus',[rnmfmethodname],ids,'postProcessing',[postprocessingsetting],'cropped', [windowingsetting]);
• Instead of rnmfmethodname, insert the method of the following methods, on which reults you want to run the segmentation on: 'robustNMF', 'robustBreg', 'robustNMF_excludeWHS' or 'robustNMF_excludeWHS_Breg'
• Replace postprocessingsetting by: true, false or both, depending on if you would like to run the Refinement Step or not (or get results for both cases)
• Replace windowingsetting by true or false, depending on if you would like to apply the windowing approach of the window calculated in the RNMF step
• Please adapt the settings for the segmentation in ./segment.m