Efficient feature selection based novel clinical decision support system for glaucoma prediction from retinal fundus...

Area of Science:

• Medical Imaging
• Machine Learning
• Ophthalmology

Background:

• Glaucoma is a progressive optic nerve disease causing irreversible vision loss, with a projected increase in cases worldwide.
• Early detection and accurate diagnosis are crucial for managing glaucoma and preventing severe visual impairment.
• Current diagnostic methods can be enhanced by advanced machine learning techniques for improved accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a novel metaheuristic-based feature selection technique for enhanced glaucoma classification.
• To apply the Gravitational Search Optimization Algorithm (GSOA) for identifying the most influential features from retinal fundus images.
• To improve the performance of machine learning models in diagnosing glaucoma using selected features.

Main Methods:

• Retinal fundus images from public and private datasets were utilized, extracting 36 features.
• The Gravitational Search Optimization Algorithm (GSOA) was employed for metaheuristic-based feature selection.
• Six machine learning models were trained and evaluated using the selected feature subsets, employing 70:30 data split, 5-fold, and 10-fold cross-validation.

Main Results:

• The proposed GSOA-based feature selection technique significantly enhanced the classification performance of machine learning models.
• The study achieved a high accuracy of 95.36% in glaucoma classification.
• Eight statistical performance metrics and execution time were calculated to validate the effectiveness of the approach.

Conclusions:

• The GSOA-based feature selection method demonstrates a promising approach for accurate glaucoma detection using retinal fundus images.
• This technique can serve as a valuable tool for medical practitioners, offering a reliable second opinion and potentially reducing the burden on healthcare professionals.
• The successful application of GSOA in feature selection holds potential for improving patient outcomes by enabling timely intervention and preventing vision loss.