Enhanced Disease Detection Using Contrast Limited Adaptive Histogram Equalization and Multi-Objective Cuckoo Search in Deep Learning

Abstract

Delayed diagnosis of numerous diseases often results in postponed treatment, adversely affecting patient outcomes.By analyzing biological signals and patient photographs, critical information about an individual's health or the severity of a medical condition can be obtained for various diseases.Signals from Electroencephalography (EEG), Electrocardiography (ECG), and Electrooculography (EOG) can be used to predict and diagnose disorders related to the brain, heart, eyes, muscles, and nervous system.Additionally, biomedical images acquired through X-ray, ultrasound, and magnetic resonance imaging can be utilized for disease diagnosis and detection with the help of image processing techniques, artificial intelligence, and deep learning methods.In this study, we propose a novel approach that combines the Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithm and Multi-Objective Cuckoo Search (MOCS) with Convolutional neural networks (CNNs) to achieve highly accurate disease classification using chest X-ray images.Our method begins by applying a contrast enhancement strategy, specifically, the CLAHE algorithm, with MOCS for optimal parameter selection to attain the highest classification performance.Subsequently, contrast-enhanced images are fed into the CNNs to further improve image quality and classification accuracy.Our approach is employed to categorize three types of chest X-ray images, namely, unhealthy, normal (healthy), and pneumonia.To assess the performance of our proposed method, we utilize the widely-used "COVID-19 Radiography" dataset.Experimental results yield an accuracy rate of 99.16%, a precision rate of 99.20%, and a sensitivity rate of 98.99%.These findings demonstrate that our proposed model outperforms existing techniques in the literature and can be effectively employed for disease detection and classification.