A Multiclassification Framework for Skin Cancer detection by the concatenation of Xception and ResNet101

Abstract

Skin cancer is a deadly type of cancer that is responsible for millions of fatalities annually across the globe. This malignant cancer occurred due to the proliferation of abnormal epidermal cells, which subsequently spread to adjacent tissues and spread to other organs and tissues through the lymph nodes. Changes in lifestyle and sun-seeking behaviors have contributed to the increase in the incidence of skin cancer. It is critical to accurately identify and classify skin cancer to prevent the serious effects that result from delaying detection and treatment. This research paper presents a newly developed deep learning model that makes use of two advanced artificial intelligence techniques, Xception and ResNet101. This method attains an extraordinarily high degree of accuracy by using the special advantages of two strong networks. The Xception-ResNet101 (X_R101) model is capable of differentiating specific categories of skin cancers, such as melanoma (Mel), melanocytic nevus (Mn), basal cell carcinoma (bcc), squamous cell carcinoma (Scc), benign keratosis (Bk), Actinic keratosis (Ak), Dermatofibroma (Df) and Vascular lesion (Vl). The implementation of borderline SMOTE improves performance substantially. A comparison is performed between the proposed methodology and four benchmark classifiers: MobileNetV2 (BM3), DenseNet201 (BM4), InceptionV3 (BM1), and ResNet50 (BM2) and state-of-the-art classifiers. To evaluate performance of the proposed methodology, three publicly available datasets (PH2, DermPK and HAM10000) are utilized. The X_R101 model attains a prediction accuracy of 98.21%. The method's accuracy and effectiveness provide benefits to dermatologists and other healthcare practitioners in terms of timely identification of skin cancer.