Deep learning model for classifying shoulder pain rehabilitation exercises using IMU sensor.
Kyuwon Lee, Jeong-Hyun Kim, Hyeon Hong and 4 others
PMID 38539223WHAT IT FOUND
A deep learning model correctly identified 11 different shoulder exercises from wrist sensor data in patients with shoulder pain.
It achieved 92.5% accuracy on new data, though it occasionally confused similar band exercises. This suggests automated monitoring is possible, but clinical validation is still needed.
Key findings
01The model achieved a test accuracy of 0.925 when classifying exercises from data it had not seen during training.
02The model frequently confused shoulder extension with band and shoulder external rotation with band, predicting one as the other.
03Shoulder abduction with a band was the easiest exercise for the model to identify, while the lawnmower exercise was the hardest.
STILL TO COME
How it was doneWhat they foundWhat it means for PTsWhat it means for OTs
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 model was trained and validated on data from the same group of patients, meaning it has not yet been tested on new patients or in different environments. The study only included patients with shoulder pain; the model's performance for other conditions or healthy individuals is unknown. The sensor was worn on the wrist, which limits its ability to distinguish exercises that produce similar wrist trajectories, such as shoulder extension and external rotation with a band. This is a proof-of-concept study for classification accuracy; it does not demonstrate that the technology improves patient outcomes or adherence in a real-world clinical setting.
Declared interests
The study was supported by non-U.S. government funding. The wearable sensor used was developed and used internally by the research team.
The easy way to misread this
Do not assume this model is ready for clinical use or that it can accurately monitor all shoulder exercises. It was tested only on data from the same patients it was trained on, and it struggles to distinguish between similar movements like shoulder extension and external rotation with a band.