Texture features variability in ultrasound video of atherosclerotic carotid plaques

Abstract

The discrimination of texture between normal and abnormal (asymptomatic or symptomatic) atherosclerotic carotid plaque in ultrasound videos is important for evaluating the gravity of the disease in subjects at risk of stroke. In this work, we present an integrated system for assessing the texture features variability in ultrasound videos of the common carotid artery (CCA). Texture features were extracted from areas around the atherosclerotic plaques and walls from ultrasound videos acquired from 30 subjects (10 normal (N), 10 asymptomatic (A) and 10 symptomatic (S)). All videos were intensity normalized prior features extraction. By identifying the cardiac cycle in each video we generate the M-mode image and estimate systolic and diastolic states. From the normalized videos, 70 different texture features were extracted and studied throughout the cardiac cycle. It is shown that: (i) the plaque gray-scale median (GSM) for the A group is statistical significantly different when compared to the GSM of S and N groups, (ii) The coefficient of variation (%CV) in the A group is higher when compared with the S and N group, (iii) similar to this trend was also the case for features entropy, GSM, standard deviation and contrast, (iv) there is a plaque feature variability per frame throughout the cardiac cycle, and (v) this variability differs between systolic and diastolic states. It is anticipated that the proposed system may aid the physician in the clinical practice in classifying between N, A and S subjects using texture features extracted from selected areas in ultrasound videos of the CCA. However, exhaustive evaluation has to be carried out with more videos and additional features.