Our project aims to assess balance in elderly individuals using data collected by Inertial Measurement Units (IMU) to predict the Berg Balance Scale (BBS) score. Data were collected from 14 participants, with sensors placed on the torso, lower back, and head, during the execution of 14 tasks derived from the BBS.
Our methodology relies on the extraction and analysis of various features such as the horizontal area representing the change in the center of pressure on the ground, the total path length, acceleration, and angular velocity from the IMUs. Additionally, we have created a new parameter, the volume that envelops the movement trace, as a new dimension in balance assessment. Logistic regression modeling was employed to determine the predictors of BBS scores among these variables extracted from the lower back sensor.
Furthermore, we also explored the impact of sensor position (i.e., the torso and head) on the associations between BBS and IMU variables, using the lower back as a reference center of mass. Moreover, by distinguishing between static and dynamic tasks, our research demonstrates how the nature of posture and movement can influence these associations. This may suggest the need to adapt assessment protocols to different everyday life contexts.
Finally, we will implement a wide range of machine learning models, including Support Vector Machine (SVM), Artificial Neural Networks (ANN), and the XGBoost model. These advanced technologies will be applied to analyze data from sensors, with particular attention paid to important evaluation matrices such as accuracy, sensitivity, specificity, and the area under the receiver operating characteristic (ROC) curve of our predictive models.
| Date | 13 Aug 2024 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Ornwipa Thamsuwan (Supervisor) |
|---|
Nkizi, Y. (Author),
Thamsuwan (Supervisor),
13 Aug 2024Student thesis: Master's thesis › Master in Engineering: Mechanical Engineering