Machine Learning Techniques for LV Wall Motion Classification Based on Spatio-temporal Profiles from Cardiac Cine MRI

In this paper, we propose an automated method to classify normal/abnormal wall motion in Left Ventricle (LV) function in cardiac cine-Magnetic Resonance Imaging (MRI). Without the need of pre-processing and by exploiting all the images of a cardiac cycle, spatio-temporal profiles are extracted from a subset of diametrical lines crossing opposites segments of the ventricular cavity. Two machine learning techniques are adapted and tested. The first one, is based on classical Support Vector Machines (SVM) and the second one that is proposed is based on dictionary learning (DL), adapted for classification in a supervised learning fashion. The experiments are evaluated based on features extracted from gray levels of the spatio-temporal profile as well as their representations in other basis such as Fourier and Wavelet domains under the assumption that the data may be sparse in one of those domains. The best classification performance has been obtained with a three-level db4 2-Dimensional wavelet transform using Fisher Discriminative Dictionary Learning as technique of classification.