基于多视图医学图像处理的COVID-19诊断算法及应用研究

Research on COVID-19 diagnosis algorithm and its application based on multi-view medical image processing

2020年伊始,一种新型冠状病毒(Coronavirus Disease 2019,COVID-19)席卷全球,世界各国面临严重的公共卫生危机。由放射科医生人工检测肺部CT影像是否存在磨玻璃密度影是重要的诊断依据,但是由于人工检测较为耗时且受制于医生主观经验,存在检测效率低下以及主观诊断误差等问题。通过研究多视图数据的特征选择策略,构建或改进新的多视图数据特征选择策略,充分利用高维数据中许多特征之间的相关性和互补性,尽可能去冗,降低计算复杂度,提高学习算法的性能。针对医学图像分割问题,采用了深度学习中的U-Net网络;针对分割标注的二维CT图像数量的有限性,创建了COVID-19分割网络;针对多视图数据的获取,提出了多视图数据特征选择策略。

At the beginning of 2020,a novel coronavirus Disease(COVID-19) is sweeping the globe,and countries around the world are facing a serious public health crisis.It is an important diagnostic basis for radiologists to manually detect the presence of ground glass density shadow in lung CT images.However,due to the time-consuming and subjective experience of doctors,manual detection has many problems,such as low detection efficiency and subjective diagnostic error.By studying the feature selection strategy of multi-view data,a new feature selection strategy of multi-view data is constructed or improved to make full use of the correlation and complementarity of many features in highdimensional data,eliminate redundancy as much as possible,reduce computational complexity and improve the performance of learning algorithm.For medical image segmentation,u-NET network in deep learning is adopted.A COVID-19 segmentation network was created for the limited number of 2d CT images segmented.A feature selection strategy for multi-view data is proposed to obtain multi-view data.