Automated Estimation of Aortic Intima-Media Thickness from Fetal Ultrasound

摘要

Intima-media thickness (aIMT) of the abdominal aorta has proven to be an early marker for atherosclerosis and cardiovascular diseases risk assessment in young adults and children. Despite recent studies have highlighted the potential usefulness of its estimation at the fetal stage from ultrasound images, this relies on error-prone and tedious manual tracing. In this study, an automated technique for aIMT estimation from fetal ultrasound images is presented and tested against manual tracing. The proposed technique is based on narrow-band level-set methods applied to the regions surrounding the aortic lumen in order to segment the portions between the blood-intima and media-adventitia interfaces and thus estimate the aIMT. This approach was tested on images acquired from 11 subject at a mean gestational age of 29 weeks. Automatically extracted aIMT values were compared to reference values manually extracted by two interpreters using Pearson's correlation coefficients, Bland-Altman and linear regression analyses. The results indicate that the accuracy of the proposed technique is comparable to that of manual tracing. As a consequence, this approach could be potentially adopted as an alternative to manual analysis for the automated estimation of aIMT.