Objective: To develop an algorithm for real-time Freezing of gait (FOG) prediction based on the online monitoring of the stepping coherence.

Background: FOG is a common motor symptom of advanced stages of Parkinson’s disease (PD) causing a transient inability to walk despite the intention to keep walking. Previous studies have shown that FOG events can be detected with high precision by analyzing gait kinematics collected through wearable motion sensors. However, until now, fewer algorithms aimed at real-time prediction of FOG.

Method: The walking coherence from 21 PD patients was extracted from signals obtained from motion sensors worn on the participant’s shanks during gait trials with FOG-triggering events (e.g., turns). This measure was chosen due to previous evidence demonstrating impaired bilateral coordination of gait (BCG) in relation to FOG, with some additional works documenting BCG occurring together with a transient functional decoupling between cortical and sub-cortical regions around freezing events, ultimately suggesting BCG might be a direct behavioral manifestation of FOG-related brain processes. Our algorithm calculated BCG via wavelet analysis (calculating the coherence of the wavelet transform of gait signals, specifically the anteroposterior angular velocity). Then the algorithm detected the time points where the legs’ coordination was strongly violated and ‘flagged’ them as FOG-alert events.

Results: We present preliminary results based on simulated real-time streaming of the collected data. Out of 103 FOG events that were present in the collected data, the algorithm predicted 96 events (sensitivity of 93.2%), on average 1.5 seconds (SD= ± 0.75 seconds) prior to the event. The algorithm often flagged also turns (i.e., in 146 out of 252 turns, i.e., 58%). However, while turning-related false positives remain a challenge for this potential FOG prediction solution, it is known that BCG is typically altered during turning, with many evidence documenting an increase in FOG occurrences around turns. Thus ‘false alarms’ in this case may not be a major limitation.

Conclusion: Overall, these results suggest this algorithm has the potential to be incorporated within a wearable system for FOG prediction, i.e., potential FOG-triggering events are identified by the algorithm, and strategies such as external auditory cueing could be delivered to avert/limit FOG occurrences.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/a-novel-real-time-algorithm-to-predict-freezing-of-gait-in-parkinsons-patients-using-wearable-mobility-sensors-data/