AI technologies, particularly large language models (LLMs) like ChatGPT, are overshadowing physiological sensors and monitors that are similarly advancing clinical knowledge and practices. For example, we developed and successfully tested a physical therapy use case based on Athos garments that measured muscle activation using Surface Electromyography (sEMG). The project, which predated the rise of LLMs, evaluated and compared the efficacy of three sessions of insurance-covered physical therapy with a recovery program based on the Athos system.
I became a lead test subject after undergoing orthoscopic meniscus surgery. The process tasked an experienced trainer with using the Athos system to assist me in performing prescribed sets of dynamic and weight-bearing exercises intended to strengthen and stabilize the injured knee. The Athos system used garments with embedded sensors to measure muscle activation intensity. The iPad application displayed two human figures (front and back) that illuminated muscles in color from blue (low activation) to red (high activation).
In my case, the system showed low muscle activation around the operated knee and high muscle activation on the opposite leg while performing exercises prescribed by my surgeon and physical therapist. My body was transferring much of the exercise workload from the operated knee to the uninjured one. My upper body conveyed a similar pattern as it shifted workloads to unburden the side with the operated knee. I could not feel, and the trainer could not consistently see the subtle changes in posture and body mechanics behind my poor exercise performance.
She then used the app to help me balance exercise workloads and muscle activation by adjusting my posture and movement mechanics while monitoring the results in real time. Lower body adjustments helped balance workload distribution and muscle activation in my upper body. We debriefed after every supervised session and worked on retraining my brain to prevent long-term problems with posture and movement mechanics. I applied what we learned to make similar adjustments during independent workouts.
The use case and technology allowed me to recover from my second surgery in about seven months, which compared favorably with the sixteen to eighteen months following the first operation. The estimates were developed by comparing post-surgery training records.
This use case (2018-19) preceded the emergence of LLM technologies like ChatGPT. We could not determine if the Athos system relied on other AI technologies like Machine Learning and Deep Learning. The company did not survive the pandemic, and our research program ended in 2019. Still, sEMG-based tools have been used in biofeedback, pain management, and muscle relaxation therapies since the 1920s[I], so we expect to see new versions in the years ahead. Reach out if you have questions: ozzie@oprhealth.com.
[i] Randy Neblett, Surface Electromyographic (SEMG) Biofeedback for Chronic Low Back Pain, May 17.2016, PMC Pub Med Central, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4934580/
Comments