Transfer learning-driven ensemble model for detection of diabetic retinopathy disease

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is an eye condition caused by diabetes, leading to vision damage and potential blindness.
• Manual diagnosis of DR from fundus images by medical experts is time-consuming and prone to error.
• Automated detection methods using computer vision are crucial for early and accurate DR diagnosis.

Purpose of the Study:

• To develop and evaluate an ensemble model for automated diabetic retinopathy detection using transfer learning.
• To improve the accuracy and efficiency of DR diagnosis compared to existing methods.
• To leverage deep learning models for analyzing retinal fundus images.

Main Methods:

• An ensemble model was developed using six deep learning convolutional neural network (CNN) models: DenseNet-169, VGG-19, ResNet101-V2, Mobilenet-V2, and Inception-V3.
• Transfer learning (TL) was employed, utilizing pre-trained models on large image datasets.
• A data-preprocessing strategy was implemented to enhance model accuracy and reduce training costs.

Main Results:

• The proposed ensemble model achieved a high accuracy of up to 98% in detecting diabetic retinopathy stages.
• The model demonstrated superior performance compared to existing approaches on the same dataset.
• The study successfully identified the stage of diabetic retinopathy using automated analysis of retinal scans.

Conclusions:

• The developed transfer learning-based ensemble model offers a highly accurate and efficient solution for automated diabetic retinopathy detection.
• This approach can aid in early diagnosis and management of DR, potentially preventing vision loss.
• The findings highlight the potential of deep learning and computer vision in advancing ophthalmic diagnostics.