Rethinking risk of falls in stroke survivors

Deviations in gait kinematics using machine learning

Shutterstock 1412024846

This project addresses two research questions: (1) Is it possible to use a reduced set of only three inertial
measurement units (IMUs) to detect responses to gait perturbation in stroke survivors? (2) Is it possible
to estimate kinematic information from IMUs?

The ability to adequately respond to perturbations is crucial to avoid falls, but this ability is limited in many
patient populations, among which stroke survivors. Researchers have evaluated this ability mainly using
parameters that quantify deviations in gait kinematics after a perturbation. Unfortunately, these
parameters do not predict falls, which is the long-term goal of these investigations. Possible reasons are:
a. responses to perturbations are highly variable,
b. lab technology influences the patient’s movements and
c. the choice of predictive parameters is subjective.

Unsupervised learning is one solution to overcome the problem of subjectivity and to quantify deviations
between unperturbed and perturbed gait. In this project, an optical marker-based dataset will be used for
this as it contains all information from a gait lab. To make the approach suitable for clinical application,
we aim to predict the same measures from three IMUs in a second step.

This will allow for data collection in a large number of patients, without the use of expensive, complex
and labour-intensive technology in a patient-friendly environment by employing gait perturbations
occurring in daily life. Finally, more data and a different view on the data will help to tailor therapeutical
efforts to the patient’s needs and consequently improve rehabilitation.

Participating organisations

Vrije Universiteit Amsterdam
University of Applied Sciences Utrecht
Netherlands eScience Center
Life Sciences
Life Sciences
Social Sciences & Humanities
Social Sciences & Humanities

Team

SD
Sina David
Lead Applicant
Vrije Universiteit Amsterdam
Michiel Punt
Michiel Punt
Sonja Georgievska
Sonja Georgievska
eScience Research Engineer
Netherlands eScience Center
Jisk Attema
Programme Manager
Netherlands eScience Center
Yang Liu
Yang Liu
eScience Research Engineer
Netherlands eScience Center
Elena Ranguelova
Elena Ranguelova
Tech Lead
Netherlands eScience Center
Cunliang Geng
eScience Research Engineer
Netherlands eScience Center

Related projects

Emotion Recognition in Dementia

Advancing technology for multimodal analysis of emotion expression in everyday life

Updated 22 months ago
Finished

Classifying activity types

Gaining insights from wearable movement sensors

Updated 18 months ago
Finished