Classification of Myocardial Infarcted Patients by Combining Shape and Motion Features

Myocardial infarction changes both the shape and motion of the heart. In this work, cardiac shape and motion features are extracted from shape models at ED and ES phases and combined to train a SVM classifier between myocardial infarcted cases and asymptomatic cases. Shape features are characterised by PCA coefficients of a shape model, whereas motion features include wall thickening and wall motion. Evaluated on the STACOM 2015 challenge dataset, the proposed method achieves a high accuracy of 97.5i¾?% for classification, which shows that shape and motion features can be useful biomarkers for myocardial infarction, which provide complementary information to late-gadolinium MR assessment.