基于深度学习的人机协同病理显微镜设计

Design of Human-Computer Collaborative Pathology Microscope Based on Deep Learning

摘要: 本文结合传统的病理光学显微镜，提出了基于深度学习进行人机协同的病理显微镜设计方法，将病理图像倍率识别模块、病理图像拍摄模块、病理图像处理模块、分割结果投影模块等嵌入传统光学显微镜。其中在图像处理模块的算法部分，我们在医学图像分割网络Unet网络基础上加入Critic模块建立深度学习模型，该模型可以同时学习全局和局部的特征，分割癌变的区域。本文提出的显微镜设计方法可以实现病理医生在物镜下移动病理样本，显微镜可以实时分割图像并展示癌变区域。

Abstract: Combining with the traditional pathological optical microscope, a pathological microscope design method based on deep learning and man-machine collaboration is proposed. The pathological image shooting module, pathological image processing module and segmentation result projection module are embedded in the optical microscope. In the algorithm part of the image processing module, the deep learning model is established by adding the CRITIC module on the basis of the medical image segmentation network UNET. The model can learn global and local features at the same time. The microscope design method proposed in this paper can realize when the pathologist moves the sample under the objective lens, the microscope can segment the image in real time and show the cancerous area.