Accelerating Balance Recovery Using Adaptive EVS Therapy

Investigators:
John Ralston, Neursantys Inc.
VP Nguyen, UMass Amherst

MassAITC Cohort: Year 3 (Aging)

Project Accomplishments: This project developed a novel machine learning-powered framework to enable personalized therapies for older adults using the NEURVESTA platform—a head-mounted, wearable sensory neuromodulation device that delivers stochastic wideband sub-threshold electrical vestibular stimulation (swsEVS) for non-invasive balance restoration. The project had two goals: developing ML algorithms to customize and adapt swsEVS stimulation parameters based on neurophysiological impairments, and analytically validating those algorithms using retrospective datasets. Using IMU signals collected from the companion Phybrata wearable sensor during repeated treatment sessions, the framework converts cleaned head sway recordings into longitudinal improvement trajectories and classifies subjects into three clinically actionable states: continued improvement, stable benefit, or reassessment for parameter modification. The system was tested on data from 32 older adults who completed an 18-session fixed-parameter protocol. 

The project achieved what the team believes to be the first ML-powered personalization framework for swsEVS stimulation using a simple ear-worn IMU sway signal. Early-decision validation demonstrated that stable benefit could be detected with high sensitivity by session 14, with a zero sham false-flag rate, and external validation using Timed Up and Go (TUG) functional mobility testing showed positive alignment between sway-derived improvement trajectories and functional outcomes. A peer-reviewed manuscript on the framework’s findings has been submitted for publication, and the team is pursuing additional funding for larger prospective studies to advance toward a truly closed-loop adaptive stimulation system. 

Initial Proposal Abstract: Neursantys’ NEURVESTA wearable bioelectronic solution is a cutting-edge platform that integrates both diagnostic sensing and therapeutic treatment for degraded sensory and motor balance functions. This project will develop ML-driven methods to adapt the treatment to each patient’s unique sensory and motor impairment profile to increase the effectiveness of NEURVESTA’s current treatment protocol.

The NEURVESTA treatment protocol includes both a pre- and post-treatment diagnostic assessment using a novel neurophysiological impairment sensing technology called physiological vibration acceleration (phybrata) that quantifies each patient’s unique balance impairment profile. The subsequent therapeutic treatment utilizes a specialized form of non-invasive transcranial electric stimulation developed by Neursantys, called sub-threshold wideband stochastic Electrical Vestibular Stimulation (swsEVS), that has been shown to induce persistent neuroplastic recovery of degraded balance in patients between 55 and 90 years of age. The current NEURVESTA platform utilizes the same fixed swsEVS parameters for all patients, and treatment consists of three 20-minute sessions per week over a 6-week period.

By harnessing a retrospective data set comprised of over 2000 pre- and post-treatment phybrata measurements and unique balance maps, we will develop advanced ML algorithms to generate adaptive adjustments of swsEVS stimulation parameters that will further enhance therapeutic effects of the NEURVESTA treatment. This personalized closed-loop approach is expected to reduce the number of treatment sessions required to attain maximum balance recovery, which will lower treatment delivery costs and improve patient accessibility and adherence to NEURVESTA balance treatments.

Outcomes: