Development of a Pain Signaling System Using Machine Learning

A wearable pain signaling system is introduced, to be used in the daily care for people with severe and profound intellectual disabilities. Due to several medical disorders, this group of people can experience daily acute pain and chronic pain, which they have trouble expressing and communicating to caregivers. The system consists of a smart sock with fabric sensors, a sensor unit sending a 6A current through the smart sock receiving electrodermal response and a mobile application containing a machine learning algorithm translating the signal. The pain signaling algorithm uses data of one to five painful stimuli from 28 healthy participants. Random forest modeling was used to classify moments of pain and train a model to predict pain from new data. The algorithm’s accuracy could be improved by an ensemble of five models and voting, so this groundbreaking system can become a much-wanted attribution to daily caregiving of people with disabilities.