Hybrid deep learning technique for optimal segmentation and classification of multi-class skin cancer

Abstract

ABSTRACT This study introduces a novel deep learning-based approach for skin cancer diagnosis and treatment planning to overcome existing limitations. The proposed system employs a series of innovative algorithms, including IQQO for preprocessing, TSSO for cancer region isolation, and FA-MFC for data dimensionality reduction. The USSL-Net DCNN extracts hidden features, and the BGR-QNN enables multi-class classification. Evaluated on Kaggle and ISIC-2019 datasets, the model achieves impressive accuracy, up to 96.458% for Kaggle and 94.238% for ISIC-2019. This hybrid deep learning technique shows great potential for improving skin cancer classification, thus enhancing diagnosis and treatment outcomes and ultimately reducing mortality rates.