医学图像精细化分割方法研究

Study on Refine Segmentation Method for Medical Images

指出了深度学习方法在医学图像分割中取得较大的进展,但医学图像处理的复杂性使得全自动分割方法难以取得较好的分割效果。在卷积网络分割的基础上,结合应用场景使用适当的后处理手段来提升图像的分割效果是一种比较有临床意义的研究方法。主要研究了基于概率图的全连接条件随机场模型和基于用户交互的深度编辑网络,并分析总结了这两种方法的实现原理和各自优势,并对未来的研究工作进行了一些展望。

Deep learning has made great progress in medical image segmentation,but the complexity of medical image processing makes it difficult for automatic segmentation to achieve better results.On the basis of convolutional segmentationnetwork,it is a clinically significant research method to use appropriate post-processing methods to improve the image segmentation effect in combination with application scenes.The paper mainly studies the full connected conditional random field model based on probability graph and the deep editing network based on user interaction.It also analyzes and summarizes the implementation principles and respective advantages of these two methods,whichgives some prospects for future research.