Multi-class learning for vessel characterisation in intravascular ultrasound

Abstract

In this thesis we tackle the problem of automatic characterization of human coronary vessels in Intravascular Ultrasound (IVUS) image modality. The basis for the whole characterization process is machine learning applied to multi-class problems, where the Error-Correcting Output Codes (ECOC) framework is used as central element for the design of multi-class classifiers. Two main contributions are presented in this thesis. First, a novel method for the design of potential function for Discriminative Random Fields is presented, namely ECOC-DRF. The method is successfully applied to problems of object classification and segmentation in synthetic and natural images. Furthermore, ECOC-DRF is applied to obtain a robust classification of the main morphological areas of coronary vessels in IVUS sequences. Based on ECOC-DRF, the main regions of the coronary artery are robustly segmented by means of a novel holistic approach, namely HoliMAb, representing the second contribution of this thesis. The HoliMAb framework is applied to problems of lumen border and media-adventitia border detection, achieving an error comparable with inter-observer variability and with state-of-the-art methods.