Predicting Dementia Based on Wearable Digital Devices Data

Abstract

Introduction

Dementia is a progressive neurodegenerative brain disorder and the most severe expression of cognitive impairment. Research has shown that preventive interventions targeting lifestyle and vascular risk factors early on can slow down the neurodegeneration process. In a previous lifestyle intervention study a prediction model based on clinical and demographic data was built, predicting a high or a low risk of developing dementia. While these data have been found to be good predictors of dementia, this study aims to utilize wearable digital devices data with the goal of using machine learning (ML) methods to predict dementia. Mobile devices and wearable digital consumer technology could help detect the disease specific changes earlier in the disease course. Cognitive, sensory, and motor changes could be detected 10-15 years prior to the diagnosis of dementia.

Methods

160 participants aged 60-77 are included in the 2-year study. Disease-relevant features are collected with a smartphone and a smartwatch. The features we will include in the prediction model are metrics which reflect movement (phone acceleration, relative location), phone usage, levels of exercise (steps, activity log, calories) and sleep metrics. From these features symptoms of dementia will be extracted based on literature review. Also, an ML prediction model with the outcome of high or low risk of developing dementia will be built. Outcomes of this model will be compared with the ones of the prediction model based on clinical and demographic data. The specific ML approach (supervised or unsupervised learning) is yet to be determined after comparative analysis of both ML techniques. We hypothesize that digital devices data alone are not enough to accurately and precisely predict whether a participant has a high or a low risk of developing dementia. However, by adding digital devices data to the already existing prediction model, its prediction accuracy and precision would likely improve.

Discussion

The goal of this project is to build a prediction model for dementia based on digital biomarkers. An important benefit of digital devices for both researchers and patients is the widespread use of phones and other digital devices in daily life, which eliminates additional expenses for collecting data. Measurements could be collected for longer periods instead of sporadically. This could improve detecting and predicting dementia early on, enabling interventions, which can slow down the disease progression.

References

[1] S. Hanke et al., “AI-based predictive modelling of the onset and progression of dementia,” Smart Cities, vol. 5, no. 2, pp. 700–714, 2022. doi:10.3390/smartcities5020036

[2] L. C. Kourtis, O. B. Regele, J. M. Wright, and G. B. Jones, “Digital biomarkers for alzheimer’s disease: The mobile/wearable devices opportunity,” npj Digital Medicine, vol. 2, no. 1, 2019. doi:10.1038/s41746-019-0084-2