Discrimination of benign- versus-malignant skin lesions by thermographic images using support vector machine classifier

摘要

Skin cancer is the cancer type with the highest incidence in the world. Its diagnosis requires a specialist physician, with expertise in skin diagnostics. Thermography is a noninvasive technique based on the detection of infrared emission that is completely safe to humans. In this study, thermal images of clinically similar lesions were registered and analyzed aiming to provide a noninvasive diagnostic information for discrimination of: basal cell carcinoma versus intradermal nevus, squamous cell carcinoma versus actinic keratosis, and melanoma versus pigmented seborrheic keratosis. Thermal images were analyzed using a MATLAB (R) routine to evaluate statistical, histogram, and filtering metrics of each image, and a support vector machine classifier was used to discriminate the lesions based on those metrics values. Actinic keratoses and squamous cell carcinoma showed distinct average temperatures, whereas the other pairs of lesions presented similar temperatures. Nevertheless, the benign lesions showed higher definition of borders detection than malignant lesions, as a general rule. The results showed that support vector machine classifier has great ability for discrimination of clinically similar lesions based on their thermal images, suggesting that the thermography can be used as an auxiliary tool for the diagnosis of skin malignant lesions. Published by AIP Publishing.