摘要
手动分割大规模心血管疾病医学图像对于医生是极其耗时繁琐的任务,所以自动分割冠状动脉计算机断层血管造影(computed tomography angiongraphy,CTA)是研究心血管疾病的重要基础。本文提出了一种基于深度学习的分割方法,冠状动脉CTA图像结合对应标签作为全卷积神经网络的训练数据实现了血管精确分割。本文所提出的方法在2种评价指标Jaccard系数与Dice系数上取得了0.763和0.834评分,能够对冠脉CTA进行准确的三维分割,为医生提供辅助诊断的作用。
Manual segmentation of medical images of large-scale cardiovascular disease is extremely time-consuming and cumbersome for doctors.Therefore,computed tomography angiography(CTA)is an important basis for studying cardiovascular diseases.In this paper,a segmentation method based on deep learning is proposed.Coronary CTA images combined with corresponding tags as training data of V-net network realize accurate segmentation of blood vessels.In the two evaluation indicators,the method proposed in this paper has repectively achieved 0.763 on the Jaccard coefficient and 0.834 on Dice coefficient,which can accurately segment the coronary CTA and provide a auxiliary diagnosis for the doctor.
作者
刘敏
方志军
高永彬
LIU Min;FANG Zhijun;GAO Yongbin(School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《智能计算机与应用》
2020年第3期72-75,共4页
Intelligent Computer and Applications
关键词
深度学习
V-net网络
冠状动脉三维分割
辅助诊断
deep learning
V-net network
three-dimensional segmentation of coronary arteries
auxiliary diagnosis
