A Comprehensive Review of Metaheuristic-Based Hyperparameter Optimization for Convolutional Neural Networks in Skin Lesion and Wrinkle Detection

The rapid evolution of deep learning has transformed dermatological image analysis, with Convolutional Neural Networks (CNNs) emerging as the dominant architecture for skin lesion classification and wrinkle detection. However, the effectiveness of CNNs is heavily constrained by the complex, high-dimensional hyperparameter space, making manual or grid-based tuning impractical. Metaheuristic optimization algorithms, including Particle Swarm Optimization (PSO), Genetic Algorithms (GA), Ant Colony Optimization (ACO), and their hybrid variants, have recently gained traction as powerful tools for automated, efficient hyperparameter search in medical imaging applications. Despite numerous individual studies demonstrating significant performance gains, no systematic review has yet synthesized and critically evaluated these approaches specifically for skin feature detection tasks. This paper presents a systematic literature review (SLR) that categorizes, compares, and analyzes metaheuristic-based hyperparameter optimization techniques applied to CNNs for skin lesion and wrinkle detection. Key findings reveal consistent improvements in mean Average Precision (mAP), robustness across skin tones, and computational efficiency, while highlighting persistent challenges in scalability and bias mitigation.

کلیدواژه ها:

Deep Learning ، Convolutional Neural Networks ، Metaheuristic Optimization ، Hyperparameter Tuning ، Skin Lesion Detection ، Wrinkle Detection ، Particle Swarm Optimization ، Genetic Algorithm ، Ant Colony Optimization