Comparison of machine learning strategies for infrared thermography of skin cancer

The aim of this work was to explore the potential infrared thermal imaging as an aiding tool for diagnosis skin cancer lesions, using artificial intelligence methods. Thermal parameters tumours were retrieved from thermograms and used input features two machine learning based strategies: ensemble deep learning. strategy outperformed one, showing good predictive performance differentiation melanoma nevi (Precision = 0.9665, Recall 0.9411, f1-score 0.9536, ROC(AUC) 0.9185) non-melanoma 0.9259, 0.8852, 0.9051, 0.901). IRT combined with techniques is promising simplifying accelerating cancer. Despite ongoing awareness campaigns cancer’ risk factors, its incidence rate has continuously been growing worldwide, becoming a major public health issue. standard first detection method – dermoscopy –, largely experience-dependent mostly assess melanocytic lesions. As innocuous technique that maps surface temperature, which may be associated pathological states, e.g., tumorous it could all conditions. application methods process collected temperature data can save time assist care professionals low experience levels in task. To best our knowledge, study where set expanded lesion approach.