Objective: To investigate the potential to predict the impact of the patient’s symptomatology on his ability to walk. Specifically, we (1) formed clusters of performance during walking based on the evaluation of patients and healthy controls; and (2) we developed and validated a multinomial regression model to predict the performance group membership based on the patient’s condition.

Background: Symptomatology of patients living with Parkinson’s disease can now be assessed using inertial measurement units (IMUs). The resolution offered by this type of assessment combined with the possibility to be performed on a regular basis without the presence of a neurologist motivated the development of innovative signal processing and analytical approaches to derive a clinical portrait.

Method: 107 patients with PD and 69 controls equipped with a full-body inertial measurement system performed a timed up and go (TUG). Performance was assessed using the time required to complete the task. K-means method was used for clustering the overall performance. Each participant also underwent a clinical evaluation of their symptoms, as well as anthropometric measures were taken. Change point analysis was used to determine at which value a specific symptom had an impact on the performance group. A multinomial regression model was then derived using 80% of the patients’ data, and validated with the remaining 20%.

Results: Preliminary results reveal the presence of three performance groups: normal TUG, those slightly affected during a TUG, and those greatly affected. The change point analysis allowed to reduce each variable to 2 levels. The statistical model revealed that a BMI greater than 25.4 increased the risk of being in the slightly affected group. The presence of freezing as well as an age greater than 76 increased the risk of being greatly affected, while bradykinesia also tended to increase that risk. The current model allowed us to classify correctly 81% of the patients, based on their symptomatology and anthropometric measures.

Conclusion: This type of approach allows to make the most out of a useful, but complex set of information. This specific case is considered a step forward towards the ability to predict, from a clinical evaluation, the impact of the personalized clinical profile on the patient’s gait.

« Back to 2019 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/predicting-the-impact-of-parkinsons-disease-on-a-patients-gait-the-potential-of-innovative-approaches-based-on-instrumented-clinical-tests/