The top 3 best combinations of 2-3 LGE features obtained from Fisher’s linear discriminant analysis (LDA) for predicting block or reentry in one more simulations for a given image.

All features were max-min normalized to the range 0-1. Coefficients are the weight values assigned by the LDA to each normalized feature. AUC values are area under the curve from a 20-fold cross validated receiver operating curve analysis whose threshold values (obtained from 1 fold) are the feature linear combinations (obtained from the remaining 19 folds).