Enhanced Gated MLP-Based Lung Disease Classification Using Grey Wolf Optimized (GWO) Multi-Feature Extraction With Cumulative Learning

Lung diseases are prevalent in today’s world, exhibiting a wide range of disease variations that requires early diagnosis for preventing disease evolvement, diminishing long-term complications, and limiting the fatality rate. Conventional feature extraction methods such as scale-invariant feature transform (SIFT), discrete wavelet transform (DWT) and classification models like convolutional neural network (CNN), multi-layer perceptron (MLP), graph neural network (GNN), are associated with high computational complexity, redundancy in feature structure, and restricted adaptability due to rigid designs. Moreover, it demonstrates limitations in scalability, generalization over limited datasets, susceptible to noise or illumination and lack of optimized feature sets leading to weak predictive performance. In order to overcome that, a novel framework that enhances lung disease prediction using Grey wolf optimization (GWO) based optimized multi-feature extraction with cumulative, recursive learning and a gated multi-layer perceptron (GMLP) for robust classification. This approach provides robustness to noise or illumination i.e. invariant feature extraction, efficient dimensionality and complexity of the extracted features, captures multi-scale spatial-textural patterns through global and local feature fusion and it provides high scalability and generalization capability while maintaining computational efficiency. Also, the optimization approach leads to progressive feature enhancement, knowledge retention, adaptability and redundancy elimination for the extracted features which are further processed by a classifier that selectively integrates full and regional patterns to sharpen feature distinctiveness and improve classification performance. The results obtained from the proposed framework tested on benchmark datasets were encouraging, showcasing notable performance with 98% accuracy outperforming the evaluated baseline methods.