Bradykinesia assessment in Parkinson’s disease

Start date: 18-08-2020
End date: 18-02-2021

Background

Bradykinesia is one of the cardinal symptoms of Parkinson’s disease (PD). Bradykinesia is defined as a slowness to initiate a movement, with a progressive reduction in speed and amplitude. This decrease in amplitude or speed, also known as sequence effect, is often seen while performing repetitive movements. The management of bradykinesia is determined by clinical evaluation using standardized clinical assessments. However, the reliability of those assessments usually falls short. In general, the problem lays in the large inter-rater variability. In the case of bradykinesia, symptoms as sequence effect, akinesia and freezing are all condensed in one score; making harder to standardize the rating process. In addition, the use of standardized scales requires an in-person clinical examination, which includes burdensome travel time for patients and provides only a partial representation (i.e. one time-point assessment) of the overall disease impairment. Objective assessments, such as the online keyboard test or the analysis of someone’s typing behaviour, may address the limitations in clinical evaluation by diminishing the inter-rate variability and allowing the detection of small variations due to long-term monitoring. We expect that the analysis of the online keyboard test data and the typing behaviour data of participants using robust data analysis techniques may reveal patterns that can be used to estimate bradykinesia in a more reliable and standardized manner.

We have collected data from 20 people with PD, 19 people with Cerebellar Ataxia and 20 healthy controls. Each of them have performed a keyboard test composed by 3 tasks. Each of those tasks were performed for 60 seconds and all keystroke times recorded. Using the keystroke times, we aim to extracted: (1) dwelltime - time a key is pressed, (2) flighttime - time the finger is above the key, and (3) tap speed - taps per second. Variation on those outcomes may also be investigated. After extraction, data analysis techniques are applied to those parameters, to its variation, and to the clinical details of participants in order to (1) to identify patterns in this large dataset that can detect the presence of sequence effect in PD and (2) determine which outcomes derived from typing and clinical data can be used in a discriminant model to classify disease status (PD vs CA vs HC).

Research questions

The aims of this study are:

(1) to identify patterns in this large dataset that can detect the presence of sequence effect in people with PD;

(2) determine which outcomes derived from typing and clinical data can be used in a discriminant model to classify disease status (PD vs CA vs HC).

Tasks

  • To learn about Parkinson’s disease and how technology may benefit clinical practice;
  • To apply robust data analysis techniques to measures the variance among data points in order to develop an algorithm for bradykinesia detection and/or assessment;
  • To test the performance of the algorithms developed in this project;
  • To evaluate the feasibility of applying machine learning techniques to a clinical dataset;
  • To implement together with AI for health department an online user interface in order to deploy the algorithm developed during this process;
  • To present the results achieved in this project during a research meeting at the Expertise Center for Parkinson’s disease and Movement Disorders at the Radobudumc.

Innovation

When the algorithm(s) developed in this study have sufficient sensitivity and accuracy, and have been externally validated, they may will be implemented into a digital web interface to allow its implementation into clinical practice and possibly monitoring of people with PD at home.

People

Anna Gansen

Anna Gansen

Master student

Artificial Intelligence, Radboud University

Ana Ligia Silva de Lima

Ana Ligia Silva de Lima

Postdoctoral researcher

Expertise Center for Parkinson’s disease and Movement Disorders, Radboudumc

Marjan Meinders

Marjan Meinders

Senior researcher

Healthcare Improvement Science, Radboudumc

Twan van Laarhoven

Twan van Laarhoven

Assistant professor

Data Science, Radboud University

Bas Bloem

Bas Bloem

Professor

Neurology, Radboudumc