Abstract

Info

Author(s):: SUN Guodong1; SHI Yunyu1; LIU Xiang1; SONG Jialin2; ZHAO Jingwen1; PU Xiuli1; YIN Ling1; 1. School of Electrical and Electronic Engineering, Shanghai University of Engineering Science, Shanghai 201620, China 2. Department of Ultrasound, Changzheng Hospital Affiliated to the Second Military Medical University, Shanghai 200003, China

Keywords:: Keywords: ultrasound image feature extraction machine learning vessel segmentation review

Abstract:: Abstract: The blood vessel segmentation algorithm for ultrasound image and its evaluation indexes are mainly reviewed. Although the classical image processing algorithms based on feature extraction cant get rid of the reliance on manual labor, it is a feasible research approach to make full use of the traditional and mature technical methods in the research scenario where there is a lack of large samples of ultrasound blood vessel images. The algorithms based on machine learning improve the generalization ability of segmentation algorithm and overcome the shortcomings of the traditional methods. However, deep learning techniques have strong dependence on data and poor interpretability, and their effectiveness and stability need to be further studied. The study on blood vessel segmentation evaluation algorithms is critical, and finding the objective evaluation methods suitable for segmentation of blood vessels in ultrasound images is one of the important topics. In conclusion, the traditional method is still an effective method to complete the blood vessel segmentation in ultrasound images, and the combination of the traditional method and deep learning techniques is the future development trend.