Sway frequencies may predict postural instability in Parkinson's disease: a novel convolutional neural network approach.
David Engel, R Stefan Greulich, Alberto Parola and 5 others
PMID 39966853WHAT IT FOUND
A computer model distinguished people with Parkinson's from healthy controls using sway data, achieving 100% accuracy in the side-to-side direction.
This pilot study involved only 18 participants, so these high scores likely reflect overfitting rather than a proven clinical test.
Key findings
01The model achieved 100% accuracy, sensitivity, and specificity when classifying side-to-side body sway data.
02The model relied on high-frequency sway components (above 1 Hz) to make its decisions, which may reflect rigidity or proprioceptive issues rather than tremor.
03The study is explicitly described as a pilot with limitations regarding generalizability due to the small and heterogeneous sample size.
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 is a pilot with a very small sample size (18 patients, 15 controls), which increases the risk of overfitting. The participants varied widely in age, disease duration, and medication dosage, limiting generalizability. All patients were on dopaminergic medication, so the results do not apply to untreated patients. The high accuracy may not translate to new, unseen patients due to the limited data pool.
Declared interests
The study was funded by the Deutsche Forschungsgemeinschaft, European Commission, Hessisches Ministerium für Wissenschaft und Kunst, and Philipps-Universität Marburg.
The easy way to misread this
Do not interpret the 100% accuracy as evidence that this test is ready for clinical use. The authors explicitly state the study is of pilot nature and cannot exclude overfitting due to the small sample size of 18 patients.