Back to Search Start Over

Construction of levodopa-response index from wearable sensors for quantifying Parkinson's disease motor functions

Authors :
Memedi, Mevludin
Thomas, Ilias
Nyholm, Dag
Westin, Jerker
Senek, Marina
Aghanavesi, Somayeh
Medvedev, Alexander
Askmark, Håkan
Aquilonius, Sten-Magnus
Bergquist, Filip
Constantinescu, Radu
Ohlsson, Fredrik
Spira, Jack
Lycke, Sara
Ericsson, Anders
Source :
Publikationer från Örebro universitet.
Publication Year :
2016
Publisher :
Örebro universitet, Handelshögskolan vid Örebro Universitet, 2016.

Abstract

The goal of this study was to investigate the feasibility of wrist worn motion sensors to objectively measure motor functions in Parkinson’s disease (PD). More specifically, the aim was to construct a sensor-based levodopa-response index (SBLRI) and evaluate its clinimetric properties (convergent validity and internal consistency). Nineteen advanced PD patients and 22 healthy controls were recruited in a single center, open label, single dose clinical trial in Sweden. The subjects performed standardized motor tasks while wearing one sensor on each wrist and one on each ankle. Each sensor unit consisted of three-dimensional accelerometer and gyroscope. The patients were video recorded and the videos were blindly rated by three independent movement disorder specialists. The clinical scores were given using the Treatment Response Scale (TRS) on a scale from -3 = ‘Very Off’ to 0 = ‘On’ to +3 = ‘Very dyskinetic’. The clinical assessments were based on the overall motor function of the patients. A mean TRS was defined as the mean of the three specialists’ assessments per time point. The measurements were repeated over several time points following a single levodopa/carbidopa morning dose (50% over normal to induce dyskinesias). Sensor measurements during rapid alternating movements of hands were processed with time series analysis methods to calculate spatiotemporal parameters designed to measure bradykinesia and dyskinesia. For each hand, 96 spatiotemporal parameters were calculated and their average scores were then used in a principal component analysis to reduce the dimensionality by retaining 6 principal components. These components were then used as predictors to support vector machines and to be mapped to the mean TRS ratings of the three specialists and to calculate the SBLRI. For this analysis, a 10-fold stratified cross-validation was performed. The SBLRI was strongly correlated to mean TRS with a Pearson correlation coefficient of 0.79 (CI: 0.74-0.83, p

Details

Language :
English
Database :
OpenAIRE
Journal :
Publikationer från Örebro universitet
Accession number :
edsair.dedup.wf.001..76a0315b56a9bbca8a7f94a6fb610a78