Toward Sensor-Based Early Diagnosis of Cognitive Impairment of Elderly Adults in Smart-Home Environments using Poisson Process Models

Abstract

Emerging sensor-based assessment in combination with machine learning methodologies provide the potential to revolutionize current practices of (early) diagnosis of dementia. The goal of this research is to detect cognitive impairment in elderly adults using sensor-based measures. Longitudinal time-series data of sensor signals are analyzed with advanced computational models and supervised machine learning algorithms to identify individuals with cognitive impairment. This research further designs novel computational models using Poisson Processes that can model subtle temporal changes in sensor-based measurements, therefore have the potential to provide more reliable descriptors of cognitive impairments compared to aggregate time-series measures. Our results indicate that the proposed approach can effectively distinguish between dementia and MCI based on the sensor features yielded by the Poisson Process. Sensor-based assessment that relies on the non-homogeneous Poisson Process is further found to be effective in differentiating between adults with dementia and healthy adults, and depicts better performance compared to expert-based assessment. Findings from this research have the potential to help detect the early onset of cognitive impairment for elderly adults, and demonstrate the ability of advanced computational models and machine learning techniques to predict one’s cognitive health, thus, contributing toward advancing aging-in-place.