A texture-based probability mapping for localisation of clinically important cardiac segments in the myocardium in cardiac magnetic resonance images from myocardial infarction patients

This paper presents a novel method for the identification of myocardial regions associated with increased risk of life threatening arrhythmia in patients with healed myocardial infarction assessed by late enhanced gadolinium magnetic resonance images. A probability mapping technique is used to create images where each pixel value corresponds to the probability of that pixel representing damaged myocardium. Cardiac segments are defined as the set of pixel positions associated with probability values between a lower and an upper threshold. From the corresponding pixels in the original images several features are calculated. The features studied here are the relative size and entropy values based on histograms with varying number of bins. Features calculated for a specific cardiac segment are compared between patients with high and low risk of arrhythmia. The results from comparing a large number of cardiac segments indicate that the entropy measure has a better localisation property compared to the relative size of the myocardial damage, and that the localisation is more focused for fewer number of bins in the entropy calculation.