Revolutionizing scalp disease detection using integrating SCSAEN classifier with secretary bird optimization for enhanced diagnostic precision

Automated recognition of scalp-related disorders is challenging because fine-grained similarities among conditions are often obscured by lighting variations, skin-tone diversity, and image noise. Traditional examination relies heavily on clinical expertise and may produce inconsistencies in complex cases. To overcome these limitations, this study proposes a deep learning system integrating contrast-sensitive enhancement, structural edge detection, hierarchical feature encoding, and adaptive classification for precise scalp analysis. Initially, dynamic contrast enhancement amplifies hidden textures and tonal variations in scalp images. Laplacian of Gaussian (LoG) filtering subsequently sharpens intensity transitions and delineates boundary regions essential for identifying subtle abnormalities. The preprocessed images are passed through a customized VGG-16 network to extract multiscale features while preserving low-level patterns and high-level structural cues. A Stacked Convolutional Sparse Autoencoder Network (SCSAEN) compresses and reconstructs these features to obtain compact, noise-resistant representations for classification. The Secretary Bird Optimization Algorithm (SBOA) fine-tunes the model parameters, improving convergence and generalization across diverse inputs. Compared with existing approaches, the proposed model achieved superior diagnostic performance, attaining 99.80% accuracy, 97.64% precision, 96.81% recall, and 98.81% specificity. It also obtained a Dice coefficient of 98.12% and completed the processing cycle within 180.4 s, demonstrating its potential for accurate and efficient clinical decision support.