A machine learning approach to carotid wall localization in a-mode ultrasound

Abstract

ARTSENS is being developed as a fully automated ultrasound based imageless system to facilitate mass screening of patients for early detection of atherosclerosis especially in low- and middle- income countries. ARTSENS uses a single element ultrasound transducer and thus makes its measurement on basis of observations on A-line. Positioning the single element transducer on the carotid artery and automatic identification of proximal and distal walls are a major challenge in this device. In this paper, we explore various machine learning methods namely – logistic regression, support vector machine and Adaboost, on selectively extracted features. The algorithms were trained on data from 60 subjects and tested on data from 40 subjects. Adaboost algorithm performed the best among the three logging a 91.66% accuracy.