Automated Extracting the Region of Common Carotid Artery Based on the Contour Shape in Ultrasound Imaging of Neoglottis

The esophageal phonation proficiency is evaluated subjectively by specialists such as doctors. Therefore, for the purpose of objective evaluation of proficiency, a measuring method of the cross-sectional area of the neoglottis lumen on ultrasound images has been proposed. However, the position and shape of tissues in the ultrasound image differs depending on the position and angle at which the probe is applied to the subject’s body during imaging. For this reason, it is not possible to compare the cross-sectional area of the neoglottis lumen for each imaging and to make an absolute evaluation. In this paper, we propose a method for extracting the common carotid artery region in ultrasound images to evaluate of the cross-sectional area of the neoglottis lumen absolutely based on the size of the common carotid artery. Then, the validity of the region extraction of the common carotid artery was verified using ultrasound videos during vocalization by the neoglottis for two laryngectomees. As the result, the average IoU of the extracted region was 0.659, the accuracy was affected to the imaging position and the clarity of the contour shape of the common carotid artery.