UNET Architecture Based Myocardium Segmentation Of Cardiac MRI
Bakal Vinita and Arbind Kumar Gupta; International Journal of Research and Analytical Reviews (IJRAR) June 2019, Volume 6, Issue 2
In this paper we present deep learning methods to automate myocardial segmentation in MRI images. There are some enhancements done to existing systems in this work. Dice metric and Jaccard Index are chosen as optimization objective functions. A batch normalization layer has been incorporated in the convolutional block to train fully convolutional neural network. Our architecture has much improved accuracy as compared to previous approaches having similar architecture.