International Journal of Computer Techniques Volume 12 Issue 3 | A COMPARATIVE ANALYSIS OF A PRIORI KNOWLEDGE-BASED BP-ANN ANDSVM FOR DIABETIC RETINOPATHY DETECTION
A Comparative Analysis of A Priori Knowledge-Based BP-ANN and SVM for Diabetic Retinopathy Detection
International Journal of Computer Techniques – Volume 12 Issue 3, May – June 2025
Abstract
Diabetic retinopathy (DR) is a leading cause of vision loss. This study analyzes a dataset of 180 fundus images to compare traditional and enhanced Backpropagation Artificial Neural Networks (BP-ANN) with Support Vector Machines (SVM) for DR detection. Feature extraction incorporated prior clinical knowledge such as vessel diameter and tortuosity. Evaluation using k-fold cross-validation confirms that the a priori knowledge-based BP-ANN model outperforms all others in early-stage DR detection.
Keywords
Diabetic Retinopathy, Fundus Images, Blood Vessel, Retinal Detection, BP-ANN, A Priori Knowledge, SVM.
Conclusion
Incorporating prior anatomical features into the BP-ANN significantly improved detection accuracy for diabetic retinopathy when compared to traditional BP and SVM models. The user-friendly GUI developed further supports practical clinical deployment, showing promise for early DR screening and timely treatment interventions.
References
- A. Abdullah (2022). “Early Detection of Diabetic Retinopathy Based on Artificial Intelligent Techniques.” IJEEI.
- M. Hardas et al. (2022). “Retinal Fundus Image Classification for DR Using SVM Predictions.” Physical and Engineering Sciences in Medicine.
- B. Jones et al. (2022). “Identification of DR Using Fuzzy Logic and BP Neural Networks.” IJCSIS.
Post Comment