Abstract
Smartphone-based digital biomarker (DB) assessments provide objective measures of daily-life tasks and thus hold the promise to improve diagnosis and monitoring of Parkinson’s disease (PD). To date, little is known about which tasks perform best for these purposes and how different confounds including comorbidities, age and sex affect their accuracy. Here we systematically assess the ability of common self-administered smartphone-based tasks to differentiate PD patients and healthy controls (HC) with and without accounting for the above confounds. Using a large cohort of PD patients and healthy volunteers acquired in the mPower study, we extracted about 700 features commonly reported in previous PD studies for gait, balance, voice and tapping tasks. We perform a series of experiments systematically assessing the effects of age, sex and comorbidities on the accuracy of the above tasks for differentiation of PD patients and HC using several machine learning algorithms. When accounting for age, sex and comorbidities, the highest balanced accuracy on hold-out data (73%) was achieved using random forest when combining all tasks followed by tapping using relevance vector machine (67%). Only moderate accuracies were achieved for other tasks (60% for balance, 56% for gait and 53% for voice data). Not accounting for the confounders consistently yielded higher accuracies of up to 77% when combining all tasks. Our results demonstrate the importance of controlling DB data for age and comorbidities.
Highlights
Diagnosis of Parkinson’s disease (PD) still often relies on inclinic visits and evaluation based on clinical judgement as well as patient and caregiver reported information
MACHINE LEARNING ALGORITHMS As a different machine learning (ML) algorithm may provide the best performance for a given task, we evaluated four commonly applied algorithms for differentiation between PD and Healthy Control (HC): 1) Least Absolute Shrinkage and Selection Operator (LASSO) is a linear method commonly used to deal with high-dimensional data
We report the following measures of predictive performance: balanced accuracy (BA), sensitivity, specificity, positive (PPV) and negative predictive value (NPV), mean receiver operating characteristic (ROC) curves with 95% confidence intervals and area under the curve (AUC)
Summary
Diagnosis of Parkinson’s disease (PD) still often relies on inclinic visits and evaluation based on clinical judgement as well as patient and caregiver reported information. Recent studies have identified digital assessments as such promising objective biomarkers for PD symptoms including bradykinesia [2], [3], freezing of gait [4], [5], impaired dexterity [6], balance and speech difficulties [7]–[9] Most of these results were obtained with a moderate number of participants and in a standardized and controlled clinical setting, reducing generalizability and limiting an interpretation with respect to applicability of these measures to an at-home self-administered setting [10]–[12]. Recent studies applying machine learning algorithms to these high-dimensional data suggested a good diagnostic sensitivity of the respective digital assessments for detection of Parkinson’s disease [14]– [17] Such at-home assessments create a range of new challenges including selection bias, confounding and sources of noise that need to be understood and dealt with to ensure good reliability of respective outcomes to a level that is sufficient for at home data collection [18]. Age, sex and comorbidities are known confounding factors
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
