Ultrasound-based automated carotid lumen diameter/stenosis measurement and its validation system

Abstract

Objective.�Degree of carotid stenosis is an important predictor to assess risk of stroke. Systolic velocity-based methods for lumen diameter and stenosis measurement are subjective. Image-based methods face a challenge because of low gradients in media and intima walls. Methods.�This article presents AtheroEdge� 2.0, a two-stage process for automated carotid lumen diameter measurement that combats the above challenges. Stage one uses spectral analysis based on the hypothesis that far-wall adventitia is brightest. Stage two uses lumen pixel region identification based on the assumption that blood flow has constant density. Using global and local processing, lumen boundaries are detected. This clinical system outputs lumen diameter along with stenosis severity index (SSI). Results.�Our database consists of institutional review board�approved 202 patients (males/ females: 155/47) left and right common carotid artery images (404 images, Toshiba scanner). Two trained neuro radiologists performed manual lumen border tracings using ImgTracer� software. The coefficient of correlation between automated and two manual readings was 0.91 and 0.92. Dice similarity and Jaccard index were 95.82%, 95.72% and 92.10%, 91.92%, respectively. The mean diameter error between automated and two manual readings was 0.27 � 0.26 and 0.26 � 0.28 mm, respectively. Precision of merit was 98.05% and 99.03% with respect to two readings. SSI showed 97% accuracy. Conclusions.�The image-based automated carotid lumen diameter and stenosis measurement system is fast, accurate, and reliable. � 2018, SAGE Publications Ltd. All rights reserved.