Journal: Scientific Reports

Article Title: Predicting persistent central serous chorioretinopathy using multiple optical coherence tomographic images by deep learning

doi: 10.1038/s41598-022-13473-x

Figure Lengend Snippet: A schematic of deep learning (DL) model training, validation, and performance assessment for prognosis of central serous chorioretinopathy (CSC) using optical coherence tomography (OCT) image sets. Baseline horizontal OCT B-scan crossing the fovea and en face images of retinal thickness, mid-retinal, ellipsoid zone (EZ), and choroidal layers were saved as a JPEG (.jpg) file from each patient and labeled as acute or persistent disease status at 6 months from baseline. The collected images underwent resizing and preprocessing of crop and contrast adjustment. Then, a deep learning model using ResNet50 architecture was trained and validated to predict the impact on the training and validation sets. Next, a performance evaluation on the test set was carried out. Visual explanation analysis was performed using a heatmap generated with gradient-weighted class activation mapping (Grad-CAM). Image sets with high accuracy were concatenated for bimodal imaging model training.

Article Snippet: The models were trained using pretrained ResNet50 architecture on MATLAB 2021b (MathWorks, Inc., Natick, MA, USA) for each image set (B-scan, retinal thickness, mid-retinal, EZ, and choroid).

Techniques: Biomarker Discovery, Tomography, Labeling, Generated, Activation Assay, Imaging