Summary: | Skin cancer is a dangerous and potentially lethal disease that is steadily increasing worldwide. Signs of skin cancer may include changes in the appearance of moles or the emergence of new spots on the skin. Early detection is crucial, as many types of skin cancer respond well to treatment when addressed in the early stages. Computer-aided diagnostic tools are employed to aid in the diagnosis of this disease. This article introduces HASCC, a hybrid algorithm implemented through a graphical user interface for skin cancer classification. The algorithm integrates image processing, feature extraction using the VGG16 algorithm with component reduction through PCA, and classification using XGBoost trained on images from the HAM10000 dataset. The hybrid algorithm was executed and tested on a Raspberry Pi 4 embedded system. HASCC was compared at both hardware and software levels with other computational intelligence methods and architectures, revealing notable improvements in terms of accuracy, ranging from 88.2% to 93.2%, with an average execution time of 250 milliseconds at low machine resource demand during the diagnostic process. Additionally, HASCC's performance was compared against previous research focused on skin cancer detection and classification. The hardware performance demonstrates that HASCC can be implemented on single-board microprocessor devices, and its software performance suggests viability for supporting the diagnosis and classification of skin cancer.
|