Fig. 3
From: Radiomics and deep learning for myocardial scar screening in hypertrophic cardiomyopathy
Initial processing of a non-gadolinium balanced steady state free precession (bSSFP) cine image for predicting myocardial scars. A Five preprocessed images (using 5 different image filters) used for radiomics feature extraction; B Five activation maps (from 5 different channels) generated by the first convolutional neural network (CNN)Â layer in the deep learning (DL)Â model