Abstract

Info

Author(s):: HE Zhen1; SHI Yunyu1; LIU Xiang1; YANG Shaoling2; NIU Jiafeng1; 1. School of Electronic and Electric Engineering, Shanghai University of Engineering Science, Shanghai 201620, China 2. Department of Ultrasound, Shanghai Eighth Peoples Hospital, Shanghai 200235, China

Keywords:: Keywords: convolutional neural network carotid plaque intima-media thickness ultrasound image

Abstract:: Abstract: A convolutional neural network-based method is proposed for the automatic detection of carotid plaques in ultrasound images. Super-resolution generative adversarial network is used to improve the quality of ultrasonic image, and Gaussian mixture model algorithm is combined with prior knowledge to automatically extract the region of interest. Finally, the automatic detection of carotid plaques is realized by convolutional neural network. Based on the data set provided by Fengxian District Central Hospital in Shanghai, the proposed method is used to automatically detect whether there is plaque in carotid artery, and finally achieves an accuracy, sensitivity and specificity of 94.11%, 96.30% and 91.67%, respectively. Experiments have proved that the results of carotid plaque detection based on convolutional neural network have a high consistency with the true values, and that the proposed method has a good robustness.