Automated Skin Cancer Detection and Classification using Cat Swarm Optimization with a Deep Learning Model

Abstract

The application of Computer Vision (CV) and image processing in the medical sector is of great significance, especially in the recognition of skin cancer using dermoscopic images. Dermoscopy denotes a non-invasive imaging system that offers clear visuals of skin cancers, allowing dermatologists to analyze and identify various features crucial for lesion assessment. Over the past few years, there has been an increasing fascination with Deep Learning (DL) applications for skin cancer recognition, with a particular focus on the impressive results achieved by Deep Neural Networks (DNNs). DL approaches, predominantly CNNs, have exhibited immense potential in automating the classification and detection of skin cancers. This study presents an Automated Skin Cancer Detection and Classification method using Cat Swarm Optimization with Deep Learning (ASCDC-CSODL). The main objective of the ASCDC-CSODL method is to enforce the DL model to recognize and classify skin tumors on dermoscopic images. In ASCDC-CSODL, Bilateral Filtering (BF) is applied for noise elimination and U-Net is employed for the segmentation process. Moreover, the ASCDC-CSODL method exploits MobileNet for the feature extraction process. The Gated Recurrent Unit (GRU) approach is used for the classification of skin cancer. Finally, the CSO algorithm alters the hyperparameter values of GRU. A wide-ranging simulation was performed to evaluate the performance of the ASCDC-CSODL model, demonstrating the significantly improved results of the ASCDC-CSODL model over other approaches.