Classification of Parkinson's disease and essential tremor based on balance and gait characteristics from wearable motion sensors via machine learning techniques: a data-driven approach.
Sanghee Moon, Hyun-Je Song, Vibhash D Sharma and 4 others
PMID 32917244WHAT IT FOUND
Machine learning models using wearable sensor data on balance and gait could distinguish Parkinson's disease from essential tremor, with neural networks performing best.
However, accuracy was moderate and the dataset heavily favored Parkinson's cases, limiting immediate clinical application.
Key findings
01Neural network models achieved the highest performance among eight machine learning algorithms for classifying Parkinson's disease versus essential tremor based on balance and gait features.
02The study dataset was highly imbalanced, with 92.5% of participants diagnosed with Parkinson's disease and only 7.5% with essential tremor.
03A synthetic minority oversampling technique (SMOTE) was used to address data imbalance and improved classification performance in most models.
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 dataset was heavily imbalanced with only 43 participants having essential tremor compared to 524 with Parkinson's disease. The study used a cross-sectional design with data from a single visit, which may not reflect typical variability in balance and gait. No healthy control group was included, so the models only distinguished between the two disorders. The analysis used only 48 of the 130 available features, selected manually by experts, which may introduce bias. The models used pre-processed data rather than raw sensor signals, potentially losing detailed movement information.
Declared interests
The study was funded by the International Essential Tremor Foundation.
The easy way to misread this
Do not interpret the 89% accuracy as evidence that this tool is ready for clinical diagnosis. The high accuracy is partly due to the large number of Parkinson's cases in the dataset, and the model's ability to correctly identify the minority essential tremor group was significantly lower.