Skin neoplasms dynamic thermal assessment

The incidence rates of skin neoplastic lesions have increased worldwide over the last decade. Its detection is currently experience dependent, causing diagnosis delay and, consequently, lesion worsening. Thus, the study of new diagnostic techniques, as Infrared Thermal Imaging (IRT), are a current need, for reducing the uncertainty associated to this process. In this work IRT images were used in conjunction with image analysis strategies and machine learning classifiers to characterize skin neoplastic lesions and, ultimately, distinguish different tumor types. The classification results show promising results in the classification task of melanoma and nevi differentiation, with an accuracy and sensitivity of 84.0% and 91.3%, respectively, using a learner based on support vector machines. The developed methodology can be introduced in clinical daily practice providing more information and tools to health professionals.