A deep learning based vascular wall motion tracking for ultrasound imaging

Cardiovascular diseases (CVDs) can be found in early stage by discovering unhealthy vascular characteristics such as arterial stiffness and aging. Vascular wall motion is related to these characteristics. A deep learning model is proposed to track the vascular wall motion in the ultrasound image. Ultrasound data were acquired using ultrasound imaging system. To train the deep learning model, data were generated from simulation by modeling the ultrasound data reflecting the geometry of the carotid artery, its motion (e.g., cardiac, respirational, and body movement), and noise on the signal. The trained deep learning model is tested on the data acquired from simulation and tissue-mimicking phantom experiment. The performance of the proposed method on the vascular wall motion tracking is compared with the traditional motion tracking algorithm using cross-correlation. The proposed model yielded better performance than the cross-correlation based method in simulation data and comparable performance in tissue-mimicking phantom experiment. This study presents the feasibility of the deep learning approach applied for the ultrasound vascular imaging application.