面向脉络膜分割的深度学习算法综述  

A Review of Deep Learning Algorithms for Choroidal Segmentation

作　　者：范兴鸿 陈湘萍[1] 刁昕龙 FAN Xinghong;CHEN Xiangping;DIAO Xinlong(The Electrical Engineering College,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州大学电气工程学院,贵州贵阳550025

出　　处：《软件工程》2025年第4期1-5,21,共6页Software Engineering

基　　金：贵州省科技计划项目(黔科合支撑[2022]一般184)。

摘　　要：脉络膜是人眼球中的一层重要组织,准确测量和分割脉络膜,对于预防、诊断以及预测视网膜疾病具有重大的医学意义。文章对光学相干断层扫描(Optical Coherence Tomography,OCT)技术在脉络膜分割中的应用进行评估,并对近年来提出的结合OCT和深度学习技术的脉络膜分割方法进行归纳和总结。研究结果显示:与传统算法相比,带有特征提取或者特征融合模块的U形网络架构有着更好的鲁棒性与泛化性。在实际应用中,该架构能有效解决OCT图像中脉络膜分割面临的边界模糊和斑点噪声等问题,已成为脉络膜分割算法中的主流模型。在眼科成像领域中,这些算法不仅提高了脉络膜分割的效率和准确性,而且还减轻了医生手工分割的工作负担,有助于提升视网膜疾病的诊断和治疗的水平。The choroid is an important layer of tissue in the human eye,and accurate measurement and segmentation of the choroid are medically significant for the prevention,diagnosis,and prediction of retinal diseases.This study evaluates the application of Optical Coherence Tomography(OCT)technology in choroidal segmentation and summarizes the choroidal segmentation methods combining OCT and deep learning techniques proposed in recent years.The results demonstrate that compared with traditional algorithms,the U-shaped network architecture integrated with feature extraction or feature fusion modules exhibits superior robustness and generalization.In practical applications,this architecture effectively addresses challenges such as boundary blurring and speckle noise in OCT images during choroidal segmentation,establishing it as the mainstream model in choroidal segmentation algorithms.In ophthalmic imaging,these algorithms not only improve the efficiency and accuracy of choroidal segmentation but also alleviate the burden of manual segmentation by clinicians,thereby enhancing the diagnostic accuracy and therapeutic efficacy for retinal diseases.

关 键 词：脉络膜分割 脉络膜厚度 OCT 深度学习 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]