Discriminating patients with paediatric idiopathic hyperhidrosis from healthy subjects with infrared thermography and machine learning classifiers

Abstract

Idiopathic hyperhidrosis (IH) is a medical condition characterised by excessive sweating beyond what is physiologically necessary for thermoregulation affecting mainly the axillae and palms. It affects seriously the quality of life of patients and has an incidence at paediatric age of 1.6%. The diagnosis is subjective relying only the the patient claim and physician perception. It is aim of this research to evaluate if dynamic infrared thermography (IRT) along with machine learning classifier on the thermal data are able to discriminate IH paediatric patients from healthy subjects. Using dynamic IRT, through convective provocation, on the views of axillae and palms, it was possible to discriminate IH paediatric patients from healthy subjects using artificial neural networks (ANN) and random forests (RF) in thermal measurements with high accuracy (>99%), the same was not possible only with the thermal data and statistics.