WCSE

Enhancing Diabetic Retinopathy Diagnosis through Symlet WaveletIntegrated Convolutional Neural Networks

Abstract— This study evaluates the enhancement of Convolutional Neural Networks (CNNs) with Symlet wavelets for classifying fundus images in diabetic retinopathy research. Symlet wavelets (sym2 through sym8)significantly improved performance metrics over a traditional CNN. Notably, the sym6 model achieved 96% accuracy and 99% precision and recall. ROC analysis revealed substantial gains, with Class 1 attaining a perfect AUC of 1.00 using Symlet-6. This adaptation demonstrated better generalization and reduced overfitting, suggesting wavelet transformations as a robust method for improving diabetic retinopathy diagnosis through enhanced image classification

Index Terms— Convolutional Neural Networks (CNN), Symlet Wavelets, Diabetic Retinopathy, Fundus Imaging, Image Classification

Songkiat Lowmunkhong
Suranaree University of Technology, THAILAND
Ratapong Onjun
Suranaree University of Technology, THAILAND
Sayan Kaennakham
Suranaree University of Technology, THAILAND
Kittirat Phattaramarut
Suranaree University of Technology, THAILAND

[Download]

Cite: Songkiat Lowmunkhong, Ratapong Onjun, Sayan Kaennakham, Kittirat Phattaramarut, "Enhancing Diabetic Retinopathy Diagnosis through Symlet WaveletIntegrated Convolutional Neural Networks," 2024 The 14th International Workshop on Computer Science and Engineering (WCSE 2024), pp. 1-5, Phuket Island, Thailand, June 19-21, 2024.