Research Article
Linking Data Collected from Mobile Phones with Symptoms Level in Parkinson’s Disease: Data Exploration of the mPower Study
@INPROCEEDINGS{10.1007/978-3-031-34586-9_29, author={Gent Ymeri and Dario Salvi and Carl Magnus Olsson}, title={Linking Data Collected from Mobile Phones with Symptoms Level in Parkinson’s Disease: Data Exploration of the mPower Study}, proceedings={Pervasive Computing Technologies for Healthcare. 16th EAI International Conference, PervasiveHealth 2022, Thessaloniki, Greece, December 12-14, 2022, Proceedings}, proceedings_a={PERVASIVEHEALTH}, year={2023}, month={6}, keywords={mobile health Parkinson’s disease mPower data}, doi={10.1007/978-3-031-34586-9_29} }- Gent Ymeri
Dario Salvi
Carl Magnus Olsson
Year: 2023
Linking Data Collected from Mobile Phones with Symptoms Level in Parkinson’s Disease: Data Exploration of the mPower Study
PERVASIVEHEALTH
Springer
DOI: 10.1007/978-3-031-34586-9_29
Abstract
Advancements in technology, such as smartphones and wearable devices, can be used for collecting movement data through embedded sensors. This paper focuses on linking Parkinson’s Disease severity with data collected from mobile phones in the mPower study. As reference for symptoms’ severity, we use the answers provided to part 2 of the standard MDS-UPDRS scale. As input variables, we use the features computed within mPower from the raw data collected during 4 phone-based activities: walking, rest, voice and finger tapping. The features are filtered in order to remove unreliable datapoints and associated to reference values. After pre-processing, 5 Machine Learning algorithms are applied for predictive analysis. We show that, notwithstanding the noise due to the data being collected in an uncontrolled manner, the regressed symptom levels are moderately to strongly correlated with the actual values (highest Pearson’s correlation = 0.6). However, the high difference between the values also implies that the regressed values can not be considered as a substitute for a conventional clinical assessment (lowest mean absolute error = 5.4).
