基于Deep learning技术的医学影像分割研究

摘要

针对传统图像分割方法对于形状起伏较大和边界模糊的目标无法获得较高分割精度的问题,文中提出使用深度学习方法分割医学图像以辅助医生进行诊断和治疗.使用卷积层、池化层和反卷积层设计了一个包含5个卷积层、3个池化层和1个反卷积层端到端的卷积神经网络,使网络可以获得与输入图像尺寸一致的输出.使用50组临床胸部CT图像的仿真与测试结果表明,所提出的深度神经分割网络能较好地实现端到端分割出胸部CT图像,网络各卷积层对不同的组织与结构均具有不同的响应值.%For traditional image segmentation methods, the problem of high undulation and blurred boundary can not achieve high segmentation accuracy. This paper proposes to use deep learning method to segment medical images to assist doctors in diagnosis and treatment. An end-to-end convolutional neural network with five convolutional layers, three pooled layers, and one deconvolution layer was designed using convolutional, pooling, and deconvolution layers to make the network available and input images Uniformly sized output. The simulation and test results using 50 sets of clinical chest CT images show that the proposed deep neural network can achieve end-to-end segmentation of chest CT images.The convolutional layers of the network have different responses to different tissues and structures value.