Eye tracking could make VR pain therapy more personal

Virtual reality is increasingly being explored as a non-pharmacological tool for managing pain. Researchers at the University of East Anglia (UEA) are now investigating whether AI and eye tracking could make these virtual experiences more responsive to individual patients. Their study suggests that changes in pupil size and blinking can provide information about cognitive workload, potentially allowing future VR therapies to adapt in real time.

The research involved 84 participants, half of whom had chronic musculoskeletal pain, including conditions such as arthritis and fibromyalgia. Participants performed memory tasks of increasing difficulty while wearing a VR headset equipped with eye-tracking technology. The research did not test whether an adaptive VR system actually reduces chronic pain. Instead, it examined an important prerequisite: whether signals from the eyes can provide a sufficiently reliable indication of how mentally demanding a VR experience is for a user.

Reading cognitive load through the eyes

Participants completed five levels of an N-back working-memory task, ranging from a baseline condition to a demanding 4-back task. As the difficulty increased, performance declined and participants reported greater mental workload. The researchers also observed larger pupil diameters and higher blink rates as cognitive demands increased. AI models were subsequently used to determine whether these eye signals could distinguish between lower and higher levels of mental workload. Models that analysed how the signals changed over time performed considerably better than approaches based on summary measurements.

Among healthy participants, the best models achieved an accuracy of between 74 and 88 percent. The results were considerably weaker among people with chronic pain, however, with accuracy ranging from 58 to 65 percent. The researchers found that age alone did not appear to explain this difference, although they could not rule out an effect.

That finding is particularly relevant for the intended application. A system designed for people with chronic pain cannot simply assume that eye responses identified in healthy users will work equally well in patients. The researchers therefore suggest that future systems may require individual calibration, learning how each person's eyes respond to changes in cognitive demand. The VR tasks themselves were generally well tolerated, with participants reporting low levels of motion sickness and discomfort.

Adaptive therapy

VR has already attracted attention as a potential tool for pain management because immersive environments can redirect attention away from physical discomfort and reduce pain-related stress. Therapeutic environments can range from calming virtual landscapes and relaxation exercises to interactive games and cognitive tasks.

Earlier research covered by us demonstrated that the way such virtual environments are designed can also influence pain responses. Researchers at Cornell University exposed 70 participants to controlled thermal pain while comparing different virtual and social settings. Participants generally tolerated higher temperatures when sharing a VR environment with another person than when experiencing VR alone. The effect was observed whether the virtual companion was someone they knew or a stranger.

That study examined experimentally induced pain rather than chronic pain, so its results cannot simply be transferred to patients with long-term pain conditions. It nevertheless illustrates that VR pain interventions may depend on more than immersion alone. Social presence, engagement and the way attention is directed can all influence the experience.

The UEA research adds another dimension: cognitive workload. Current VR applications often provide essentially the same experience regardless of how demanding it becomes for an individual user. Someone experiencing mental overload may disengage, while an experience that is too easy may not provide sufficient engagement. Lead researcher Dr Jordan Tsigarides sees an opportunity for systems that continuously monitor mental effort and adjust the virtual experience accordingly. A future headset could, for example, reduce task difficulty when signs of overload emerge or increase the challenge when a user is insufficiently engaged.

Personalisation challenge

Such a closed-loop approach would move VR from a largely predefined intervention towards therapy that responds continuously to the user. Eye tracking is particularly interesting because the sensors required are already integrated into some commercially available VR headsets, potentially allowing physiological information to be collected without additional wearable equipment. The concept could eventually extend beyond pain management. According to the researchers, similar approaches might be investigated for anxiety and depression, or outside healthcare for adaptive education and professional training.

For chronic pain, however, the current findings also expose the main obstacle. The AI models worked best in the group for whom the technology is not primarily intended: healthy volunteers. The researchers explicitly conclude that performance in participants with chronic pain is currently insufficient for reliable individual-level assessment. Patient-specific models and prospective validation will therefore be necessary before eye tracking can safely control adaptive therapy.

The study also does not demonstrate that adapting VR to cognitive load produces better pain relief, reduces medication use or improves daily functioning. Larger studies will be required to establish whether the technical ability to estimate mental workload translates into meaningful clinical outcomes. For now, the research provides a building block rather than a finished treatment. Combined with earlier evidence showing that factors such as social interaction can change pain tolerance in virtual environments, it suggests that the future of therapeutic VR may lie not simply in creating more immersive worlds, but in making those worlds respond more intelligently to the person experiencing them.