Soft Computing Based Discriminator Model for Eye Disease

Abstract

This paper presents a novel approach for glaucoma diagnosis using a lightweight convolutional neural network (CNN) architecture optimized for efficiency and accuracy. The model, designed to run effectively on resource-constrained devices, leverages depthwise separable convolutions and advanced image preprocessing to detect glaucomatous features in fundus images. Evaluated on benchmark datasets, the proposed method demonstrates high diagnostic accuracy and computational efficiency, indicating its potential for integration into real-time, accessible glaucoma screening tools, especially in underserved regions.

Introduction

Glaucoma is a progressive, irreversible eye disease and the second leading cause of vision loss worldwide, expected to affect over 111 million people by 2040. Early detection is crucial due to its asymptomatic onset. This study presents a soft computing-based system, called the Discriminator Model for Glaucoma Diagnosis (DMGD), which automates glaucoma detection from fundus images using advanced image preprocessing, feature extraction, and Support Vector Machine (SVM) classification.

Key Components of the Study

1. Problem Statement

• Manual retinal image analysis is time-consuming, subjective, and often inconsistent.

• Variability in image quality and anatomical complexity challenges accurate and scalable diagnosis.

• There is a need for automated, accurate, and robust diagnostic systems to aid early detection.

2. Objectives

• Improve diagnostic accuracy for glaucoma.

• Support ophthalmologists in clinical decision-making.

• Handle data variability and enhance scalability of retinal screening.

3. Methodology

The DMGD framework operates in three main stages:

A. Preprocessing

• Enhances image quality and isolates the optic disc (OD).

• Steps include:

  • Color channel separation and cropping (green channel highlights OD).

  • SWFCM clustering for OD segmentation.

  • Gaussian Derivative Filters for vessel detection.

  • Inpainting algorithms to remove vessels and clean the region of interest.

B. Feature Extraction

• Extracts color-based statistics (mean, standard deviation, skewness, kurtosis).

• Uses Local Binary Patterns (LBP) and LAWS texture features to capture spatial and frequency-based textures.

C. Classification

• Applies SVM with four kernel functions: Linear (LK), Polynomial (PK), Quadratic (QK), and Radial Basis Function (RBF).

• The RBF-SVM achieved the best performance in distinguishing between glaucomatous and normal cases.

4. Results and Evaluation

• Databases used: HRF, DRISHTI-GS1, RIM-ONE, ORIGA.

• Evaluation via 10-fold cross-validation.

A. Best Performance: RBF-SVM

• Accuracy: Up to 98.47% (RIM-ONE), 97.03% (DRISHTI-GS1).

• Specificity: 100% (HRF dataset).

• Sensitivity: 93.33%.

• Outperformed other kernels, especially PK-SVM, which had <75% accuracy.

B. ROC and Confusion Matrix

• ROC curves confirmed RBF-SVM’s superior classification ability (high AUC).

• Confusion matrices showed strong precision and balance in predictions.

Conclusion

This study introduces a soft computing-based Discriminator Model for Glaucoma Diagnosis (DMGD) that effectively combines advanced image preprocessing, multi-domain feature extraction, and kernel-based SVM classification. By accurately segmenting the region of interest and removing irrelevant structures, the system enhances the reliability of glaucoma detection. Among various kernel classifiers tested, the RBF-SVM consistently demonstrated the highest accuracy across multiple fundus image datasets. The results indicate that the DMGD system offers a promising tool for supporting clinical diagnosis and could serve as a reliable second opinion in glaucoma screening.

References

[1] Y. C. Tham, X. Li, T. Y. Wong, H. A. Quigley, T. Aung et al., “Global prevalence of glaucoma and projections of glaucoma burden through 2040: A systematic review and meta-analysis,” Ophthalmology, vol. 121, no. 11, pp. 2081–2090, 2014. [2] Kishore and N. P. Ananthamoorthy, “Glaucoma classification based on intra-class and extra-class discriminative correlation and consensus ensemble classifier,” Genomics, vol. 112, no. 5, pp. 3089–3096, 2020. [3] S. Murugan, A. Bhardwaj and T. R. Ganeshbabu, “Object recognition based on empirical wavelet transform,” International Journal of MC Square Scientific Research, vol. 7, no. 1, pp. 74–80, 2015. [4] N. A. Mohamed, M. A. Zulkifley, W. M. D. W. Zaki and A. Hussain, “An automated glaucoma screening system using cup-to-disc ratio via simple linear iterative clustering super pixel approach,” Biomedical Signal Processing and Control, vol. 53, no. 11, pp. 1–17, 2019. [5] D. Selvathi, N. B. Prakash, V. Gomathi and G. R. Hemalakshmi, “Fundus image classification using wavelet-based features in detection of glaucoma,” Biomedical and Pharmacology Journal, vol. 11, no. 2, pp. 795–805, 2018. [6] T. R. Kaus, V. P. Gopi, K. A. Wahid, W. Doma and S. I. Niwas, “Combination of clinical and multi-resolution features for glaucoma detection and its classification using fundus images,” Bio Cybernetics and Biomedical Engineering, vol. 38, no. 2, pp. 329–341, 2018. [7] M. Alagirisamy, “Micro statistical descriptors for glaucoma diagnosis using neural networks,” International Journal of Advances in Signal and Image Sciences, vol. 7, no. 1, pp. 1–10, 2021. [8] G. G. Jerith and P. N. Kumar, “Analysis of shearlet system based features to characterize fundus images for glaucoma diagnosis,” Journal of Medical Imaging and Health Informatics, vol. 11, no. 9, pp. 2450–2456, 2021. [9] K. G. Kannan and T. R. Babu, “Semi supervised generative adversarial network for automated glaucoma diagnosis with stacked discriminator models,” Journal of Medical Imaging and Health Informatics, vol. 11, no. 5, pp. 1334–1340, 2021. [10] U. Raghavendra, H. Fujita, S. V. Bhandary, A. Gudigar, J. H. Tan et al., “Deep convolution neural network for accurate diagnosis of glaucoma using digital fundus images,” Information Sciences, vol. 441, pp. 41–49, 2018. [11] L. Li, M. Xu, H. Liu, Y. Li, X. Wang et al., “A large-scale database and a CNN model for attention-based glaucoma detection,” IEEE Transactions on Medical Imaging, vol. 39, no. 2, pp. 413–424, 2019. [12] J. J. Gómez-Valverde, A. Antón, G. Fatti, B. Liefers, A. Herranz et al., “Automatic glaucoma classification using color fundus images based on convolutional neural networks and transfer learning,” Biomedical Optics Express, vol. 10, no. 2, pp. 892–913, 2019.

Copyright

Copyright © 2025 Dhana K, Nishanthi M, Hemalatha R, Saranya M. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.