Integrated Model for Deep Learning and Machine Learning for The Identification of Skin Cancer

Abstract

Skin cancer is a serious and potentially life-threatening condition, making early and accurate diagnosis essential for effective treatment. The most common types of skin cancer are basal cell carcinoma (BCC), squamous cell carcinoma (SCC), and melanoma. Dermatologists rely on skin images to identify the type and severity of these cancers. In recent years, researchers have been developing computer-aided diagnosis systems to assist in early detection. This paper explores a hybrid model that combines machine learning and deep learning techniques to enhance the diagnostic accuracy of skin cancer. Specifically, Convolutional Neural Networks (CNNs) are employed to analyze dermatological images, extracting detailed features that indicate various types of skin cancer. Alongside CNNs, the FractalNet model—a specialized deep learning technique—is used for diagnosis. These models are then integrated with the XGBoost algorithm in parallel. The final diagnosis decision is made through majority voting, ensuring a robust system that can identify both simple and complex features. This hybrid approach, leveraging the different strengths of CNN, FractalNet, and XGBoost, results in a more generalized diagnostic system. The HAM10000 dataset, which includes 10,015 dermatoscopic images of pigmented skin lesions, is used for model training. Data augmentation is applied to balance the dataset. For performance evaluation, precision, recall, F1-score, accuracy, and specificity are measured, with an achieved accuracy of 98.69%..