Adaptive cluster-based superpixel segmentation and BMWMMBO-based DCNN classification for glaucoma detection

Glaucoma is a chronic disease called the silent thief of sight because it has no symptoms. When glaucoma is not identified at an early stage, it results in permanent blindness. The main objective of this research work based on the evaluation of retinal fundus images for the diagnosis of glaucoma efficiently improves the performance of glaucoma detection through higher accuracy at minimum false-positive prediction by performing preprocessing, segmentation, feature extraction, and classification. This goal is achieved through the creation of a laptop application-based computer-aided detection (CAD) system for medical professionals. The proposed CAD system would provide a second opinion as a decision made by human professionals in a controlled setting, assisting the ophthalmologist in the identification of ocular disorders. In this research work, the initial step is image acquisition; here, the input images are collected from public and in-house clinical fundus images. The next step image enhancement is performed using bilateral filter with adaptive unsharp masking (BFAUM) technique, which can eradicate the noise discern in the input images. In the subsequent step, the adaptive cluster-based superpixel segmentation (ACBSS) algorithm is implemented on an enhanced image sequence for retina blood vessel extraction and then the features are extracted using gray-level co-occurrence matrix (GLCM). Ultimately, the fundus images (public ORIGA and in-house clinical fundus images) are classified by utilizing a BMWMMBO-based DCNN classifier, the accuracy is 97.9%, and 4 FP for 699 images are predicted. The combined dataset (SCES, SINDI, and in-house clinical database) achieves an accuracy of 98.4% and 31 FP for 4501 images.