AI-based video analysis for the assessment of upper limb function in children with unilateral cerebral palsy: feasibility of remote monitoring.
Youngsub Hwang, Hakje Yoo, Minkyung Kim and 2 others
PMID 41776622WHAT IT FOUND
An AI model accurately classified upper limb impairment severity in children with unilateral cerebral palsy from routine handheld clinic videos.
The tool achieved high discrimination across all four assessment indices, suggesting it could support remote monitoring and reduce clinician scoring burden.
Key findings
01The AI model achieved an overall mean area under the curve of 0.890 for classifying severity across four upper limb function indices, with dexterity showing the highest performance.
02Agreement between AI predictions and manual therapist scoring was moderate-to-good for dexterity and moderate for range of motion, accuracy, and fluency.
03False positive and negative classifications were often caused by misleading visual cues such as therapist occlusion or poor limb visibility rather than actual motor deficits.
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 included only 19 children from a single center, which limits how widely the results apply. The AI was trained to classify severity as only mild or severe, losing the detailed information of the original 0-3 or 0-4 scoring scales. No separate external test set was used; performance was measured using the same data the model was trained on via cross-validation. Video quality varied significantly due to handheld recording, and the model made errors when the therapist or clothing blocked the view of the child's arm. The study was retrospective, meaning it analyzed past records rather than testing the tool in real-time clinical use.
Declared interests
The study was funded by the National Research Foundation of Korea and the Future Medicine 20×30 Project of the Samsung Medical Center.
The easy way to misread this
Do not assume this AI tool replaces detailed clinical assessment or can accurately score every item. It currently only distinguishes between mild and severe impairment categories and its accuracy drops significantly if the child's arm is obscured by the therapist or clothing during recording.
Summarised by AI from the full paper, without a clinician reviewing it. Check it against the source before it changes what you do. Read it on PubMed →