University of Twente Student Theses


Towards digital biomarkers to assess autonomic dysfunction in Parkinson’s Disease: Feasibility of home-based photoplethysmographic signals for heart rate analysis

Veldkamp, K.I. (2023) Towards digital biomarkers to assess autonomic dysfunction in Parkinson’s Disease: Feasibility of home-based photoplethysmographic signals for heart rate analysis.

[img] PDF
Abstract:Recent evidence suggests autonomic dysfunction plays an important role in the pathogenesis of Parkinson’s disease (PD). Only a few (subjective) clinical tests assess disease progression and are used to evaluate the entire spectrum of autonomic dysregulation in patients with PD. Long-term heart rate (HR) monitoring by wrist-worn photoplethysmography (PPG) sensors could enable patient-tailored follow-up in PD patients. We aimed to develop models to assess data quality, analyzed whether several factors could influence the data quality and performed an exploratory HR analysis in relation to the symptoms of autonomic dysfunction. From an prospective PD cohort study, we included all patients with a complete one-year follow up and analyzable photoplethysmography (PPG). We divided the PPG data into one-minute epochs and annotated epochs of a stratified subset of PD patients into different categories of signal quality. We extracted 18 time and frequency characteristics from the epochs and optimized machine learning classifiers to assess data quality for two analysis of HR patterns: HR analysis and heart rate variability (HRV) analysis. We analyzed differences in the proportion of eligible data over time based on longitudinal assessment over three different weeks (Week 0 – Week 1 – Week 52). The influence of PD-specific motor symptoms was studied using tertiles based on the UPDRS symptom scores. Resting HR and maximum HR, obtained from eligible data in Week 0, were compared in relation to the gold standard in scoring non-motor symptoms, the UPDRS Part 1b. In total, PPG data of 20 PD patients were used to annotate, train, test and validate the two models for HR analysis and HRV analysis. We obtained balanced accuracies > 93% for HR analysis and > 94% for HRV analysis. We used PPG data of 431 subjects for longitudinal analysis and PPG data of 484 subjects for analyzing PD-specific factors. We did not find significant differences between time and PD-specific factors. Using 486 subjects, we showed an increase in resting HR during the day and night with increasing UPDRS Part 1b scores. The maximum HR decreased during the day whereas it increased during the night. This thesis showed that we could develop high-discriminative models to assess PPG data quality. In the exploratory HR analysis, using one of these models, we showed promising first results in linking HR parameters to the severity of autonomic dysfunction. This research should be further elaborated by focusing on longitudinal HR and HRV analysis during the night to enable patient-tailored follow-up.
Item Type:Essay (Master)
Radboudumc, Nijmegen, Netherlands
Faculty:TNW: Science and Technology
Subject:44 medicine
Programme:Technical Medicine MSc (60033)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page