Enhanced Skin Cancer Detection Using a Hybrid Deep Learning and Particle Swarm Optimization Framework

Using a hybrid technique that combines Particle Swarm Optimization (PSO) and Deep Convolutional Neural Networks (DCNNs) to improve diagnostic sensitivity and accuracy of skin cancer detection is presented in this study. During pre-processing, the mean filter is utilized for reducing noise and improve quality of images. Then, in dermoscopic images, skin lesions are precisely separated from the background using Otsu's segmentation approach. PSO is used for feature selection after segmentation, which optimizes the selection process by determining the most pertinent characteristics that provide precise classification. By reducing the number of dimensions in the data, this phase enhances the effectiveness and sensitivity of the classification process. In the final step, a DCNN is employed for classification using the optimal features selected by PSO. With the use of its deep architecture, the DCNN is able to accurately discern between benign and malignant skin lesions by automatically learning intricate patterns and discriminative properties. The use of PSO guarantees the DCNN's maximum performance by optimizing its hyperparameters and augmenting its generalization capabilities. The accuracy of the DCNN is 92% and the sensitivity between the classifiers of the DCNN is 90% respectively.