基于U-Net神经网络眼底图像的视盘分割研究

眼底图像中的视盘在青光眼筛查和诊断中起着重要作用。因此,从眼底图像中对视盘进行准确、快速地定位与分割具有重要意义。在过去,研究者们已经进行了对视盘的深入研究,但如何提高定位准确率和分割精度仍是视盘分割的一大难题。对此本文提出一种采用深度学习结构U-Net的视网膜视盘自动分割的方法,该方法结合机器学习,通过深度网络提取输入图像的视盘特征,从而得出相应的分割结果图。相对于传统的视盘分割方法,本文的U-Net神经网络能够有效学习有利于分割视盘的特征,从而提高分割的精确度,而且分割耗时更短。

Research on Optic Disc Segmentation Based on U-Net Neural Network Fundus Image

The optic disc in the fundus image plays an important role in glaucoma screening and diagnosis.Therefore,it is important to accurately and quickly locate and segment the optic disc from the fundus image.In the past,researchers have conducted in-depth research on the optic disc,but how to improve the positioning accuracy and segmentation accuracy is still a major problem in optic disc segmentation.To this end,this paper proposes a method of automatic segmentation of the retina optic disc using the deep learning structure U-Net.This method combines machine learning to extract the optic disc features of the input image through the deep network to obtain the corresponding segmentation result map.Compared with the traditional video disc segmentation method,the U-Net neural network in this paper can effectively learn the features that are beneficial to segment the video disc,thereby improving the accuracy of segmentation,and the segmentation takes less time.