Executive Summary
Early diagnosis of glaucoma enables an ophthalmologist to provide more effective treatment and control the progression of the illness. Patients notice the symptoms when the damage is significant; therefore, much research has been conducted on the diagnostic methods. There are several systems which recognize glaucoma, and doctors usually determine which one is more suitable for the patient after an examination. Optic Nerve Imaging is a technique which helps to detect optic nerve changes over time through documented images. The paper offers an even more effective approach to determine glaucoma by combining image-based and disease-related elements. The features of both methods were enhanced, which improved the analysis of the Optic Disc and Optic Cup texture.
Introduction
Glaucoma is an incurable eye disease affecting eyesight significantly by damaging the optic nerve head. High eye pressure occurs as a result of drainage canals being blocked. Warning signs and symptoms appear later in the disease when the optic nerve is already destroyed. A timely diagnosis of the illness helps doctors provide effective treatment and prevent blindness, as the eyesight cannot be recovered once glaucoma progresses further. Al-Akhras et al. (2019) have proposed an integrated and accurate system to diagnose glaucoma through automated disease detection. The purpose of their paper is to offer a more efficient method of diagnosing the disease by implementing both image-based and disease-related features. The authors claim that the combination of those elements will provide more information and resources to determine the condition (Al-Akhras et al., 2019). The study provides an integrated and innovative approach to diagnosing glaucoma more efficiently and promptly.
Main body
The method of the work conducted by Al-Akhras et al. consists of three major parts: image-preprocessing, extraction of features, and classification of elements. Retinal fundus images were collected for experimentation from three hospitals. Different techniques were implemented to obtain features; optic disc and optic cup were obtained by the methods proposed by other researchers (Al-Akhras et al., 2019). Image-based features were enhanced by adopting both the red and the green image channels. Disease-based features were also improved, which detects the optic cup and the optic disk segments better. Support Vector Machines (SVM) and Artificial Neural Networks (ANN) algorithms were used for developing an automatic diagnosis system and detecting the images with glaucoma. In the light of the reported methodology, the proposed techniques were successful in detecting high percentages of specificity and accuracy.
The technique distinguished thirteen images out of one hundred and six samples to be with glaucoma. Ophthalmologists proved the results to be accurate and corresponding to the real diagnoses. SVM and ANN successfully differentiated glaucoma by the use of specificity and sensitivity. As a result, with SVM, when the analysis was done according to both features without data normalization, the highest classification accuracy was 87.74%, and the specificity was 100% using the Gamma value of 0.25. The sensitivity value of 23.08% was reached with the help of the Gamma value of 1. With data normalization, the highest classification accuracy of 80.19% and specificity of 91.40% were retrieved using a Gamma value of 0.25. The Gamma value of 0.8 showed the highest sensitivity value of 30.77. The ANN accuracy of 98% was reached both with and without data normalization (Al-Akhras et al., 2019). The outcome indicated that the combination of the proposed features proved to be more effective when utilized simultaneously.
Further enhancements are recommended for detecting glaucoma through automated diagnosis. The data should be documented and collected by hospitals, educational and research centers to provide enough information for groundwork. The study was limited because of the short supply of the existing database. The accuracy of the data is not entirely reliable, as the technologies used to detect glaucoma cannot recognize faulty information entry during the examination and diagnosing. Minute errors in segmentation can also cause misinterpretations and affect the diagnose outcome. The technological and informational limitations occurring during the experiment enquire more improvements in the sphere of glaucoma automated diagnosing.
Conclusion
The work proposed an advanced methodology for diagnosing glaucoma in the earliest stages to provide better treatment and avoid the damaging consequences of the disease. Image-based and disease-related features were offered to be combined in screening the patients. An automatic diagnosis system was developed by SVM and ANN, and sample images were classified into normal and those showing the signs of glaucoma. Automated screening detects early changes in the images during regular checks, which can be taken by the patients themselves, learning the right angle and using the system. Further improvements are necessary for the areas of enough database availability and technology to obtain effective outcomes.
Reference
Al-Akhras, M., Barakat, A., Alawairdhi, M., & Habib, M. (2019). Using soft computing techniques to diagnose glaucoma disease. Journal of Infection and Public Health, 1-8. Web.