Towards the development of a walking assist-as-needed protocol for patients with Parkinson's disease : detection of freezing of gait episodes in real time

Details

Supervisors

Ronsse, Renaud

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil biomédical, à finalité spécialisée

Abstract

The number of patients suffering from Parkinson's disease continues to increase, primarily due to aging population but also to a growing exposure to environmental factors such as pesticides. The symptom of freezing of gait, affecting approximately half of individuals with Parkinson's disease, is highly disabling and disrupts the walking ability of these patients. Treating this symptom is challenging, and traditional drug therapies such as levodopa does not always yield positive outcomes. Consequently, several studies, although somewhat limited, have investigated the potential of robotic gait assistance as a means to alleviate this disabling condition. The objective of this work is to determine how freezing of gait episodes can be detected in real-time in order to enable future on-demand robotic assistance, therefore providing assistance only when needed. The effectiveness of three implementation methods (Freeze Index, surface electromyography-based method and K-index) is tested by comparing the results obtained with those reported in articles presenting these methods. This comparison is conducted in two ways. The first approach involves testing the algorithms developed on publicly available data. The second one involves testing the algorithms developed on data obtained experimentally as part of this study. The obtained results concerning sensitivity and specificity diverged to some extent from the existing literature, displaying relatively inferior performances. Drawing conclusions about the superiority of one algorithm over another is challenging due to the variability between results across different datasets. Indeed, with the open access dataset, the K-index gives better performances followed by the Freeze Index. It is therefore the surface electromyography-based method which seems to be the weakest at detecting FOG episodes. On the other hand, testing the algorithms on the experimental data gives another ranking. The surface electromyography-based method becomes the better one, followed by the K-index and the Freeze Index. Although optimum performances are not achieved, this work effectively demonstrates the feasibility of developing a freezing of gait episode detection algorithm. It elucidates the influence of the studied algorithm parameters on performances and offers potential ideas for enhancement.