Automatic Malignant and Benign Skin Cancer Classification Using a Hybrid Deep Learning Approach

• DOI: 10.3390/diagnostics12102472
• Publication Date: 2022-10-13T00:53:29Z
• Authors: Atheer Bassel; Amjed Basil Abdul...; Zaid Abdi Alkaree...; Nor Samsiah Sani; Husam Jasim Mohammed
• Keywords: Medicine (General); skin cancer; deep learning; CNN; machine learning; prediction; skin cancer; deep learning; prediction; 02 engineering and technology; Article; 3. Good health; machine learning; R5-920; 0202 electrical engineering, electronic engineering, information engineering; CNN

Skin cancer is one of the major types with an increasing incidence in recent decades. The source skin arises various dermatologic disorders. classified into based on texture, color, morphological features, and structure. conventional approach for identification needs time money predicted results. Currently, medical science utilizing tools digital technology classification cancer. machine learning-based robust dominant automatic methods classifying existing proposed deep neural network, support vector (SVM), network (NN), random forest (RF), K-nearest neighbor are used malignant benign identification. In this study, a method was stacking classifiers three folds towards melanoma cancers. system trained 1000 images categories benign. training testing were performed using 70 30 percent overall data set, respectively. primary feature extraction conducted Resnet50, Xception, VGG16 methods. accuracy, F1 scores, AUC, sensitivity metrics performance evaluation. Stacked CV method, levels by learning, SVM, RF, NN, KNN, logistic regression Xception techniques achieved 90.9% accuracy stronger compared to ResNet50 VGG 16 improvement optimization large dataset could provide reliable system.