Genetic and Etiological Insights from Automated Lumen Diameter Measurements in Carotid Ultrasounds of the UK Biobank

Abstract

Carotid ultrasound is routinely used in clinical practice for non-invasive vascular anatomical and functional assessment. In particular, the carotid intima-media thickness (cIMT) is an important marker for quantifying atherosclerotic burden in the common carotid arteries (CCAs). Recent research suggests that several risk factors associated with higher cIMT, such as high blood pressure, can induce a compensatory increase in the carotid lumen diameter (cLD) of the CCAs. However, the genetic architecture of cLD and its association with other cardiovascular traits are still poorly understood. To address this gap, we trained a deep learning model to segment the carotid lumen from ultrasound images and developed an algorithm to measure the mean cLD. After quality control, we derived this trait for 18 808 genotyped individuals from the UK Biobank, comparing multiple measures of cLD from lateral and central views of the left and right carotid arteries. We then performed genome-wide association studies (GWAS) for cIMT, cLD and their ratio, revealing that cLD has an estimated heritability of 31%, substantially higher than that of cIMT (23%) and cIMT/cLD (14%). Finally, we conducted association and survival analyses and found that cLD was more strongly associated with several cardiovascular risk factors, such as hypertension and BMI, while cIMT/cLD performed the best in a prognostic setting after adjusting for demographics and common cardiovascular risk factors.