Deep learning approach in ultrasound image segmentation for patients with carotid artery disease

Abstract

Precision medicine is one of the most promising approaches in combating diseases that have not been treated effectively so far. The basic premise is that every patient with the observed disease is unique and that traditional “one-size-fits-all” treatments benefit only a minority. One of the most common diseases in the human cardiovascular system is carotid artery stenosis (CAS). Early detection of this disease is very important because if it is not adequately treated, it may potentially have deteriorating consequences, such as a debilitating stroke. In order to determine the correct patient-specific diagnosis that also considers the individual anatomy of the particular patient, it is necessary to perform simulations using a patient-specific geometry. This goal is achievable only if all boundaries of the areas of interest, lumen and wall, are detected. The image segmentation task can be efficiently solved by using deep learning techniques, which facilitates the process of extracting the coordinates. Then, these coordinates should be used for the further computer-based 3D reconstruction of the entire patient carotid artery and the blood flow simulations. A presented approach can be used to address the needs for stratified and personalized therapeutic interventions in the current era. Additionally, the identification and classification of the atherosclerotic plaque components such as lipid core, fibrous, and calcified tissue are presented. This task is essential to preestimate the risk of atherosclerosis progression and to stratify patients as a high or low risk.