Detection of Low Back Physiotherapy Exercises With Inertial Sensors and Machine Learning: Algorithm Development and Validation.
Abdalrahman Alfakir, Colin Arrowsmith, David Burns and 3 others
PMID 35998014WHAT IT FOUND
A machine learning system using three wearable sensors correctly identified McKenzie exercises and good sitting posture in healthy people.
It does not yet work for patients with low back pain or those performing exercises incorrectly.
Key findings
01Three sensors placed on the lower back, thigh, and ankle provided the best performance for classifying exercises and posture, achieving high accuracy scores.
02The system was trained and tested only on healthy individuals with no history of low back pain who performed the exercises correctly.
03Convolutional neural networks required longer recording segments to accurately classify dynamic exercises, which may be difficult for patients with pain.
STILL TO COME
How it was doneWhat they foundWhat it means for PTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study included only 19 healthy participants with no history of low back pain, so it is unknown if the system works for patients with pain or limited movement. Participants performed exercises correctly under supervision; the system has not been tested on incorrect or poor-quality technique. The dataset was small due to COVID-19 restrictions, which may limit the generalizability of the machine learning models. The study did not test the system's ability to distinguish between 'good' and 'forced good' posture, as the models could not separate them. The CNN model required long recording segments (10-12 seconds), which may not be feasible for patients who cannot hold positions or perform movements for that duration due to pain.
Declared interests
Several authors have financial ties to Halterix Corporation, a digital physiotherapy company. One author is a cofounder and holds equity, another works part-time for the company, and two others hold equity.
The easy way to misread this
Do not assume this system can currently monitor your patients' adherence or technique at home. It was validated only in healthy people performing exercises correctly in a lab setting, and has not been tested in individuals with low back pain or on their actual home performance.