aBnormal motION capture In aCute Stroke (BIONICS): A Low-Cost Tele-Evaluation Tool for Automated Assessment of Upper Extremity Function in Stroke Patients.
Syed A Zamin, Kaichen Tang, Emily A Stevens and 6 others
PMID 37592867WHAT IT FOUND
A machine learning tool using smartphone video scored 16 of 33 Fugl-Meyer items with 78.1% to 82.7% accuracy compared to expert raters.
This shows a low-cost way to assess arm function remotely, though it cannot yet measure all movements or strength.
Key findings
01The automated system successfully scored 16 of the 33 Fugl-Meyer assessment items using video data.
02Item-wise prediction accuracy ranged from 78.1% to 82.7%, with the dilated CNN model achieving the highest average accuracy of 82.7%.
03The tool cannot assess movements involving rotation of the shoulder or strength, limiting its clinical scope compared to a full in-person exam.
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 study was a feasibility test with a small sample size of 45 patients, so the accuracy rates may not hold up in a larger, more diverse population. The 2D video analysis could not detect shoulder rotation or hand strength, meaning it misses key aspects of upper extremity function that require physical contact or 3D depth sensing. Some items were excluded from analysis due to severe imbalance in the data (e.g., very few patients scored low on tremor items), so the model's ability to detect subtle impairments is unproven. The study was conducted in an inpatient setting with controlled camera placement; performance in a patient's home with variable lighting and angles remains untested.
Declared interests
The authors declare no conflicts of interest. The work was supported by NIH grants.
The easy way to misread this
Do not assume this tool provides a complete Fugl-Meyer assessment. It only scores 16 of 33 items and cannot assess strength, reflexes, or shoulder rotation, so a low score from this system does not capture the full extent of a patient's impairment.