Generalizing-Cardiac-Image-Segmentation-

Building Machine Learning models that generalize cardiac image segmentation using various MRI scans collected from different clinical centres. Based on Multi-Centre, Multi-Vendor & Multi-Disease Cardiac Image Segmentation Challenge (M&Ms)- https://www.ub.edu/mnms/ Cardiac MRI scans from two different vendors were taken : 'A' and 'B'. Trained a U-Net model on the scans from the vendors A and B and noted he accuracy. Concept: Using Fourier Transform, convert scans from vendor 'A'(source images) into style of scans of vendor 'B'(target images). Then trained the model on scans of 'B' + the fourier transformed images.