| 引用本文: |
穆珺,晏峻峰,彭清华.基于小波变换和U-Net的眼底图像血管分割[J].湖南中医药大学学报,2022,42(12):2052-2058[点击复制] |
|
| |
|
|
| 本文已被：浏览 1131次 下载 732次 |
| 基于小波变换和U-Net的眼底图像血管分割 |
| 穆珺,晏峻峰,彭清华 |
| (湖南中医药大学, 湖南 长沙 410208) |
| 摘要: |
| 眼底血管在许多疾病的检测和诊断中含有丰富的信息,因此提高眼底血管分割精度对于多种疾病的诊断和临床研究具有重要意义。本文提出了一种基于小波变换和U-Net神经网络的眼底图像血管分割算法,该算法通过数据增广与预处理、小波系数图像获取、U-Net神经网络训练与预测、阈值分割与后处理等步骤完成眼底血管的分割。采用分割准确率、特异度、灵敏度与AUC等指标,对本文算法进行了性能分析。实验结果显示,本文提出的眼底图像血管分割算法能得到满意的分割结果,在中医现代目诊中具有潜在的应用价值。 |
| 关键词: 眼底血管分割 哈尔小波变换 U-Net 数据增广 图像预处理 阈值分割与后处理 |
| DOI：10.3969/j.issn.1674-070X.2022.12.015 |
| 投稿时间：2022-03-25 |
| 基金项目:湖南省中医药科技计划重点项目(201901); 湖南省教育厅科学研究项目(21C0224,21C0227); 湖南省财政厅科研资助项目(2019103); 湖南省中医学国内一流建设学科建设项目(湘教通[2018]469号) |
|
| Blood vessel segmentation of retinal images based on wavelet transformation and U-Net |
| MU Jun,YAN Junfeng,PENG Qinghua |
| (Hunan University of Chinese Medicine, Changsha, Hunan 410208, China) |
| Abstract: |
| The retinal blood vessel is informative to the detection and diagnosis of various retinal diseases. Therefore,improving the accuracy of blood vessel segmentation is highly important to diagnoses of certain diseases and clinical research. A blood vessel segmentation method of retinal images based on wavelet transformation and U-Net is proposed in this paper. The blood vessel is segmented by the major phases including data augmentation and image pre-processing, wavelet coefficient images obtaining, U-Net model training and predicting, threshold segmentation and post-processing. The performance of the proposed method has been analyzed in terms of accuracy, specificity, sensitivity, AUC, and other indicators. The experimental results demonstrate that the proposed retinal blood vessel segmentation method can achieve satisfying segmentation result, and have potential application in eye diagnoses. |
| Key words: retinal blood vessel segmentation Haar wavelet transformation U-Net data augmentation image pre-processing threshold segmentation and post-processing |
|
二维码（扫一下试试看！） |
|
|
|
|
