Automatic classification of melanocytic skin tumors based on hyperparameters optimized by cross-validation using support vector machines

摘要

Melanocytic lesions may occur in various areas of the skin and may eventually develop into malignant tissue types as aresult of abnormal tissue growth. Although the gold standard for the diagnosis of melanoma is still a histopathologicalexamination, dermatologists often use dermoscopic examination in their routine practice to reduce unnecessary excisionsor to prevent misdiagnosis of clinically suspected melanocytic lesions. However, dermoscopic examinations may requirespecial training and experience. Furthermore, even among experts, different evaluation results may occur. For thesereasons, image processing and artificial intelligence application studies are performed on dermoscopic images based oninformation technologies developed in recent years. This study investigated the automatic classification of superficialspreading melanoma and nevocellular nevus using support vector machines. A publicly available and histopathologicallyverified MED-NODE data set (70 superficial spreading melanomas and 100 nevocellular naevi) was used. For theclassification task, first, the energy distributions (power spectral densities) of each image in the spectral domain wereobtained. Second, gray-level co-occurrence matrices were created, and the textural features of the matrices wereextracted. Finally, the learning model was developed with these features as input for classification. Support vectormachines were trained using validation methods, including holdout validation and stratified cross-validation. Thehyperparameters were optimized using the regularization factor of 10, the radial basis kernel function, and the gammafactor of 0.0098. Using 10-fold cross-validation, we achieved a mean accuracy of 98.9% (+/- 0.01 standard deviation),99.4% sensitivity, and 97.5% specificity.