Classification of heat-induced pain using physiological signals

Objectively assessing the pain level of a patient is crucial in various medical situations. Until now the gold standard is represented by questionnaires which have different drawbacks. To continuously assess pain, questionnaires must be answered repeatedly which is time consuming for medical stuff and prone to errors. Thus, pain automatic classification systems could improve health care, especially when patients are unable to communicate their pain level. Previous works based on heat-based pain induction predominantly tried to predict the applied temperature stimuli itself. In contrast, our work is presenting an approach to predict self-reported pain as well. Therefore, a small dataset of 10 subjects was acquired using a thermode to induce pain. Subjects were asked to rate their pain perception on a computerised visual analogue scale (CoVAS). Different classifiers were trained using both temperature and CoVAS labels. Our experiments showed the superiority of the CoVAS labels for pain recognition.