Improvement of an eye disease detection model by using the denoising diffusion implicit model.

Area of Science:

• Artificial Intelligence in Medicine
• Ophthalmology
• Medical Imaging

Background:

• Generative AI offers solutions for data scarcity in various fields, including medicine.
• Eye diseases like retinopathy, optic neuropathy, glaucoma, and macular degeneration are increasing, particularly at younger ages.
• Acquiring sufficient retinal images for training diagnostic models can be challenging.

Purpose of the Study:

• To evaluate the effectiveness of generative AI for augmenting data in eye disease prediction.
• To improve the accuracy of a convolutional neural network (CNN) model for detecting eye diseases.
• To leverage the denoising diffusion implicit model (DDIM) for generating high-quality medical image data.

Main Methods:

• Utilized the denoising diffusion implicit model (DDIM), a generative AI model, for data augmentation.
• Applied DDIM to generate synthetic retinal images to supplement limited real-world data.
• Trained a convolutional neural network (CNN) model for eye disease detection using both original and augmented datasets.

Main Results:

• The CNN model trained with DDIM-generated data achieved a 3% higher accuracy compared to training with original data alone.
• The DDIM demonstrated high inference speed and consistent generation of high-quality samples.
• The enhanced CNN model shows potential for early screening of eye disease symptoms.

Conclusions:

• Generative AI, specifically DDIM, can effectively augment medical datasets for improved diagnostic model accuracy.
• AI-driven data augmentation is a viable strategy to address data limitations in eye disease detection.
• Early detection of eye diseases through AI-enhanced models can lead to timely treatment and better patient outcomes.