A Proposed MetaNet Deep Learning for Medical Diseases Classification
International journal of intelligent engineering and systems · 2024
The Metaheuristic Gray Wolf Optimization (GWO) algorithm, based on Swarm Optimization, is widely used in various fields, including machine learning and numerical optimization.However, extracting meaningful features and useful content from large collections of medical images remains challenging for computer vision applications, necessitating improved feature classification accuracy.This paper proposes a hybrid Deep Neural Network (DNN) and GWO approach, termed MetaNet, for medical image classification.The GWO algorithm is employed to extract the most relevant features, thereby minimizing classification errors, while the DNN classifier is trained to find local minimums as part of the GWO search agent's fitness calculation.Through extensive experimentation on benchmark medical datasets, including MRI brain tumors and IRMA's ImageCLEFmed X-ray images, MetaNet demonstrated superior performance.Notably, the proposed approach achieved a classification accuracy of 98.3% on the brain tumor MRI dataset and 96.7% on the ImageCLEFmed dataset.Various performance measures such as classification accuracy, loss rate, sensitivity, specificity, recall, precision, F1 score, and ROC curve analysis were used to evaluate the system's performance.The results indicate that our proposed MetaNet model significantly outperforms traditional convolutional networks and other state-of-the-art classifiers, providing a robust and efficient method for medical image disease prediction.