Related Experiment Video
Updated: Aug 17, 2025
07:24
Home-Based Monitor for Gait and Activity Analysis
Published on: August 8, 2019
6.8K
Semi-supervised clustering of quaternion time series: Application to gait analysis in multiple sclerosis using motion
Pierre Drouin1,2, Aymeric Stamm1, Laurent Chevreuil2
1Laboratoire de Mathématiques Jean Leray, Université de Nantes, Nantes, France.
Statistics in Medicine
|December 12, 2022
Summary
Wearable motion sensors offer valuable gait analysis data for clinicians. A new semi-supervised clustering method effectively groups patients with similar walking impairments, aiding in multiple sclerosis assessment.
Area of Science:
- Biomedical Engineering
- Data Science
- Neurology
Background:
- Wearable motion sensors are increasingly used for gait analysis, providing spatio-temporal parameters.
- Analyzing complex motion sensor data requires advanced statistical methods for clinical relevance.
- Assessing gait impairment is crucial for managing neurological conditions like multiple sclerosis.
Purpose of the Study:
- To develop and apply a novel semi-supervised clustering method for gait analysis using wearable motion sensor data.
- To group patients with multiple sclerosis based on gait impairment and disease severity.
- To evaluate the effectiveness of a generalized compromise-based clustering method (hclustcompro) adapted for quaternion time series.
Main Methods:
- Utilized time series of unit quaternions derived from hip rotation data for gait measurement.
- Generalized the hclustcompro method for unit quaternion time series, incorporating quaternion dynamic time warping.
- Applied semi-supervised clustering to group multiple sclerosis patients based on gait data and clinical disability scores.
- Compared the compromise-based clustering approach with the collaborative clustering method, mergeTrees.
Main Results:
- Demonstrated the utility of wearable motion sensors for assessing gait impairment in multiple sclerosis patients.
- Successfully formed patient groups with similar walking deficiencies using the novel clustering approach.
- Showcased the benefit of integrating prior clinical knowledge to guide the clustering process.
- Found compromise-based clustering to be more appropriate than mergeTrees for this specific application.
Conclusions:
- Wearable motion sensors are effective tools for gait impairment assessment in clinical settings.
- Semi-supervised clustering, particularly the generalized hclustcompro method, is a suitable approach for analyzing complex gait data.
- Incorporating prior knowledge enhances the clinical relevance of gait analysis and patient stratification.
