AUTOMATTED CNN ARCHITECTURE BASED ON IMPROVED PSO EMPWERED WITH OBL FOR SKIN CANCER CLASSIFICATION

Abstract

Early and accurate detection of skin cancer, particularly melanoma, is critical for effective treatment and improved patient outcomes. Convolutional Neural Networks (CNN) have shown promising performance in medical image analysis, however, designing an optimal CNN architecture is complex and time-consuming. This paper proposes an automated CNN architecture for skin cancer classification based on input data by improving Particle Swarm Optimization with mutation operator and Opposition-Based Learning (OBL). The HAM10000 dataset, which comprises 10,015 dermatoscopic images, was used to evaluate the proposed model. The hybrid MPSO-OBL optimized CNN was compared with several baseline models, including PSO-CNN, GA-CNN, LeNet, and Alex-Net. The results indicate that the proposed model outperforms the baseline models across all performance metrics, achieving an accuracy of 89.5%, The proposed technique also demonstrated efficient training time, and a relatively low number of parameters compared to Alex-Net, making it suitable for deployment in resource constrained environments. The study concludes that the integration of MPSO and OBL offers a robust and computationally efficient method for automating the CNN architecture design for skin cancer classification, paving the way for future advancements in medical image analysis.