Augmented Reality System Allows Clinicians to ‘See’ Patient Pain

Researchers at the University of Michigan have developed a system that allows clinicians to “see” patient pain in real time. The technology could potentially be very useful in objectively measuring pain and identifying it in patients who have trouble communicating their symptoms.

Consisting of augmented reality glasses for the clinician and a cap containing brain sensors for the patient, the system uses artificial intelligence to convert brain activity data into a pain signature visualization so that the clinician can assess pain levels.

At present, it is difficult to measure pain levels accurately. Also, for many, such as children or patients with emotional challenges, it is sometimes nearly impossible to accurately describe their pain. “It’s very hard for us to measure and express our pain, including its expectation and associated anxiety,” said Alex DaSilva, a researcher involved in the study. “Right now, we have a one to 10 rating system, but that’s far from a reliable and objective pain measurement.”

This new technology aims to provide a visual depiction of brain activity associated with pain, as a way for clinicians to assess pain levels in their patients. Called CLARAi (clinical augmented reality and artificial intelligence), the system involves patients wearing a sensor-loaded cap, which measures changes in oxygenation and blood flow to assess brain activity in response to pain.

These data are then interpreted by an artificial intelligence system that can learn which brain signals are linked to the pain response. A clinician can view the results using augmented reality glasses, where red and blue dots indicate the location and intensity of the pain signal in a virtual brain.

The researchers trained the AI system using 21 dental patient volunteers in whom they elicited a pain response by applying cold to their teeth. The researchers found that their system could predict the presence or absence of pain approximately 70% of the time. However, with a larger training dataset these results could likely be improved.