Related Experiment Video
Updated: Nov 4, 2025
12:28
Quantification of Vascular Parameters in Whole Mount Retinas of Mice with Non-Proliferative and Proliferative Retinopathies
Published on: March 12, 2022
3.9K
A Multimodal Imaging-Based Deep Learning Model for Detecting Treatment-Requiring Retinal Vascular Diseases: Model
Eugene Yu-Chuan Kang1,2, Ling Yeung2,3, Yi-Lun Lee4
1Department of Ophthalmology, Chang Gung Memorial Hospital, Linkou Medical Center, Taoyuan, Taiwan.
JMIR Medical Informatics
|May 31, 2021
Summary
A new deep learning model can detect vision-threatening retinal vascular diseases like diabetic macular edema (DME) and age-related macular degeneration (nAMD) using multimodal imaging. This AI tool shows high accuracy in identifying diseases that require treatment.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Retinal vascular diseases, including diabetic macular edema (DME), neovascular age-related macular degeneration (nAMD), myopic choroidal neovascularization (mCNV), and retinal vein occlusions (BRVO/CRVO), are leading causes of vision loss.
- Accurate diagnosis necessitates expert interpretation of multimodal imaging data.
Purpose of the Study:
- To develop and evaluate a deep learning (DL) model for detecting various vision-threatening retinal vascular diseases.
- To assess the model's capability in identifying diseases that require treatment using multimodal ophthalmic imaging.
Main Methods:
- A retrospective study utilizing multimodal imaging data (fundus photography, OCT, FA/ICGA) from 2185 patients across three Taiwanese hospitals (2013-2019).
- A DL model was trained and validated to detect DME, nAMD, mCNV, BRVO, and CRVO, and to identify treatment-requiring cases.
- Model performance was quantified using the area under the receiver operating characteristic curve (AUC).
Main Results:
- The DL model achieved high AUCs for detecting specific diseases: mCNV (0.996), DME (0.995), nAMD (0.990), CRVO (0.988), and BRVO (0.959).
- The model demonstrated strong performance in identifying treatment-requiring retinal diseases with an AUC of 0.969.
- Heat map analysis confirmed the model's ability to pinpoint relevant retinal features for disease detection.
Conclusions:
- A deep learning model was successfully developed for the detection of multiple retinal vascular diseases.
- The model exhibits significant potential for accurately identifying treatment-requiring retinal vascular conditions through multimodal imaging analysis.
