Semantic-Segmentation-of-MRI-cardiac-images-using-the-U-NET-architecture

Project Description

In this project, a neural network is used to segment the left ventricle which can save a lot of time and increase the number of patients a doctor can diagnose compared with the traditional manual segmentation. The U-NET segmentation architecture is used for both binary segmentation and multiclass-segmentation of heart MRI images. This project tries to replicate the results of the paper ”U-Net: Convolutional Networks for Biomedical Image Segmentation”. While the original paper performs segmentation on Electron Microscope images of cells in the Drosophila first instar larva ventral nerve cor, we are using the technique to segment cardiac images instead. To achieve the A grade, we also plan to make the network perform multi-segmentation, to separately detect the left ventricle’s two components, the right ventricle, and the background.