Deep-learning-based Segmentation of Fundus Photographs to Detect Central Serous Chorioretinopathy

https://doi.org/10.1167/tvst.11.2.22
Simple Code Implementation for Deep Learning–Based Segmentation to Evaluate Central Serous Chorioretinopathy in Fundus Photography, TVST, 2022

We developed a pix2pix deep learning model for segmentation of subretinal fluid area in fundus photographs to detect central serous chorioretinopathy (CSC).

The total dataset included fundus photographs and a segmentation image dataset from 194 eyes with CSC from the medical centers and publicly accessible datasets. Additionally, we recruited 93 fundus photographs of the healthy eyes from the same center to build a classification model to discriminate CSC from normal retina.

Manual segmentation is a tedious and time-consuming task that requires domain-specific knowledge. Initially, one ophthalmologist manually screened the fundus photography images with CSC. We asked three ophthalmologists including two licensed ophthalmologists (grader 1 & 2) and one ophthalmology resident (grader 3) to segment the entire SRF area in the retinal images.

First, we prepared the dataset in Google Drive. Second, the users can upload the code file in Google Drive and open the file on the Google Drive page on the web browser. Third, prepare the dataset in Google Drive and match the folder location of the code with the address of the actual folders. In our experiment as an example, we saved the training dataset at "csc/segmentation/train/" and the test dataset at "csc/segmentation/test/" in our own Google Drive. Fourth, click the play button to the left of the code cell one by one. The second code cell links the datasets to this Colaboratory notebook using Google Drive.

The manuscript of the study was accepted to TVST, ARVO journal.

https://doi.org/10.1167/tvst.11.2.22 《基于深度学习分割的简易代码实现，用于评估眼底摄影中的中心性浆液性脉络膜视网膜病变》，TVST，2022

我们开发了一款pix2pix深度学习模型，用于分割眼底照片中的视网膜下积液（subretinal fluid, SRF）区域，以检测中心性浆液性脉络膜视网膜病变（central serous chorioretinopathy, CSC）。

本研究的总数据集包含来自医疗中心的194例CSC患眼的眼底照片及分割图像数据集，同时纳入了来自同一中心的93例健康眼眼底照片，用于构建分类模型以区分CSC患眼与正常视网膜。

手动分割是一项繁琐且耗时的工作，需要具备专业领域知识。最初，由一名眼科医生对CSC相关眼底摄影图像进行初步筛查；随后邀请3名眼科医师（包括2名执业眼科医师，即分级者1与分级者2，以及1名眼科住院医师，即分级者3）对视网膜图像中的全部SRF区域进行手动分割。

具体操作流程如下：
1. 首先在谷歌云端硬盘（Google Drive）中完成数据集的筹备；
2. 用户可将代码文件上传至谷歌云端硬盘，并通过网页浏览器在谷歌云端硬盘页面中打开该文件；
3. 于谷歌云端硬盘中完成数据集筹备，并将代码的文件夹路径与实际文件夹的地址进行匹配。本实验示例中，我们将训练数据集保存至自有谷歌云端硬盘的"csc/segmentation/train/"路径，测试数据集保存至"csc/segmentation/test/"路径；
4. 依次点击各代码单元格左侧的播放按钮。第二个代码单元格将通过谷歌云端硬盘将数据集关联至本Colaboratory（Colab）笔记本。

本研究的论文已被ARVO旗下期刊TVST收录。