부산대학교 도서관

Skin cancer is a prevalent and potentially life-threatening disease with rising global incidence rates. Early detection significantly improves prognosis, and the integration of advanced technologies in the medical field holds the promise of enhancing diagnostic accuracy. Visual inspection, which is used by dermatologists to identify skin cancer, is subjective, requires a lot of time, and is reliant on the dermatologist's ability. Diagnostics might be revolutionized by artificial intelligence technology, which offers automated, data-driven insights that increase diagnostic precision and speed. The idea that CNNs and LSTMs may be integrated to provide a dynamic and all-encompassing approach to skin cancer diagnosis served as the basis for this paper. Given sequential data, LSTMs perform better at recognizing temporal patterns, while CNNs are more adept at deriving intricate spatial features from still images. By merging these several designs, we developed a multimodal model system that can both recognize the visual characteristics of skin lesions and show how these lesions evolve over time. To compare the fusion model, first we applied CNN for skin cancer detection and achieved 90.2%. Later, we applied fusion of CNN and LSTM and acquired accuracy of 93.4%. The results shown that integrating CNN and LSTM given more accuracy for skin cancer detection.