3D Reconstruction of the Carotid Artery from Handheld Ultrasound Videos

Abhishikta Bandyopadhyay, Omkar Kulkarni, Tim Oates

Abstract: Cardiovascular disease is one of the leading causes of death among people of all genders and races in the United States. According to the CDC, approximately 695,000 people died in the United States in 2021 due to poor cardiovascular conditions and approximately 17% of these deaths were due to a stroke. To increase patients' awareness and understanding of their cardiovascular health, this paper presents a method for extracting a 3D model of the carotid artery from videos taken with handheld ultrasound devices which, due to their relatively low cost, make it possible for primary care physicians to own and use them on a large number of patients. Technicians using the device move it up and down the neck, changing positions and angles frequently. Our approach extracts the artery from each frame using machine vision methods, reorders the frames spatially (from the bottom of the artery to the top) as opposed to the temporal order in the video, and then uses computer graphics methods to build a 3D model. The method used for reordering the frames yielded an accuracy of 86.34% when compared to the ground truth sequence. Our goal is to have an easily understandable representation of the state of the carotid artery to educate patients about their risks and thereby increase compliance with treatment.