Abstract
Purpose :
This study aims to develop a multimodal deep learning model for predicting the therapeutic response to anti-vascular endothelial growth factor (VEGF) treatment in patients with diabetic macular edema (DME) based on optical coherence tomography (OCT) images and clinical data.
Methods :
A cohort of 158 DME patients who underwent three consecutive injections of anti-VEGF were included in this study. Spectral-domain OCT scans, encompassing central retinal thickness (CRT), and various clinical parameters were utilized for training the multimodal deep learning model. The model's performance was assessed through 5-fold cross-validation. Responsive patients were defined as those exhibiting a reduction of at least 50 μm in CRT after the three injections of anti-VEGF. The predictive accuracy, sensitivity, specificity, precision, and area under the receiver operating characteristic curve (AUC) were calculated to evaluate the system's performance.
Results :
The developed algorithm demonstrated an average AUC of 0.872 in discriminating responsive from non-responsive patients. The average accuracy, sensitivity, specificity, and precision were 75.2%, 90.8%, 50.0%, and 75.2%, respectively.
Conclusions :
The multimodal deep learning system accurately predicts the treatment response to anti-VEGF based on OCT images and clinical data in DME patients. The algorithm holds potential for enhancing counseling on the correction of systemic factors associated with DME and may assist in personalized therapeutic decision-making, contributing to optimized treatment outcomes.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.