Prediction of Long-Term Treatment Outcomes for Diabetic Macular Edema Using a Generative Adversarial Network
ABSTRACT
The study aimed to use generative adversarial networks (GANs) to analyze optical coherence tomography (OCT) images for predicting diabetic macular edema outcomes after long-term treatment.
METHODS
Diabetic macular edema (DME) eyes (n = 327) from a randomized controlled trial underwent anti-vascular endothelial growth factor (VEGF) treatments every 4 weeks for 52 weeks. GAN models were trained to generate probable OCT images after treatment using various combinations of input images.
RESULTS
Various GAN models showed positive predictive value, sensitivity, specificity, and kappa for residual fluid and hard exudate after long-term anti-VEGF treatment of DME. Models trained with additional input images at weeks 4 and 12 showed improved performance.
CONCLUSIONS
GAN models have potential to predict residual fluid and hard exudate after long-term anti-VEGF treatment of DME.
TRANSLATIONAL RELEVANCE
Implementation of GAN models may aid in identifying potential nonresponders after long-term treatment, thereby facilitating management planning for these eyes.
Clinical Trials and AI in Healthcare
Clinical trials are vital for developing safe and effective treatments. Our AI-driven platform, DocSym, consolidates ICD-11 standards, clinical protocols, and research, providing clinicians with an easily accessible knowledge base.
In today’s healthcare environment, streamlining operations is crucial. Our mobile apps support scheduling, treatment monitoring, and telemedicine, making it easier to manage patient care and expand services digitally.
By using AI, clinics can enhance workflows, improve patient outcomes, and reduce paperwork. Learn more about how we can help at aidevmd.com.
