Automatic Detection and Classification of Diabetic Retinopathy Using the Improved Pooling Function in the Convolution...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of blindness in diabetic patients.
• Early diagnosis of DR is crucial for effective treatment and vision preservation.
• Deep learning offers potential for automated and efficient DR screening.

Purpose of the Study:

• To develop an improved Convolutional Neural Network (CNN) model for automated DR diagnosis.
• To enhance the accuracy and efficiency of DR detection from fundus images.
• To reduce computational complexity and processing time in DR classification.

Main Methods:

• An enhanced ResNet-50 model incorporating an improved pooling function and activation function was utilized.
• The model was applied to retinal fundus images for autonomous lesion detection.
• Training and testing were performed on two publicly available datasets: APTOS and Kaggle.

Main Results:

• The proposed enhanced ResNet-50 model achieved high accuracy in DR detection.
• Accuracy rates of 98.32% on the APTOS dataset and 98.71% on the Kaggle dataset were recorded.
• The model demonstrated superior performance compared to existing state-of-the-art methods.

Conclusions:

• The developed deep learning model shows significant promise for accurate and efficient automated DR diagnosis.
• The improved pooling and activation functions contribute to reduced loss and processing time.
• This approach can aid in early DR detection, potentially preventing blindness in diabetic individuals.