Major neglected tropical diseases (NTDs) can cause skin lesions, leading to increased isolation and stigma among patients. In fact, these skin lesions are often the initial symptoms noticed by patients, even before internal organ or systemic changes occur. The World Health Organization (WHO), through its global health watch, has identified 15 NTDs and aims to eliminate them by 2030 in alignment with MDG target 3.3. Early diagnosis is crucial for achieving this goal, and our work contributes to this objective. With the rapid advancements in machine learning, computer-based diagnosis has made significant progress, particularly in the field of biomedical image diagnosis. In this project, we utilize tools such as artificial neural networks optimized by evolutionary algorithms, including the whale optimization algorithm (WOA), the shark smell optimization (SSO), and a hybrid of these two algorithms (WOA-SSA). These algorithms are initialized by a chaotic map to enhance the identification of skin diseases from clinical images. To train our neural network, we extract relevant features from the images, specifically the gray level co-occurrence matrix (GLCM) features. We then select the most reliable features and apply our optimization algorithm to improve classification accuracy while minimizing mean square error (MSE) and processing time for early and real-time diagnosis. The database used for this project comprises data from hospitals in Cameroon, as well as the Xiangya Hospital of Central South University in China. Our method achieves a global accuracy rate of 95% with the Chao-WOA-SSO hybrid optimization approach, even under varying conditions. Moreover, it demonstrates shorter computation time compared to previous methods. Therefore, it represents a promising solution for expedient and accurate diagnostics.