Early Detection of Mild Cognitive Impairment Progression Using Non-Wearable Sensor Data – a Deep Learning Approach

Abstract

Mild Cognitive Impairment (MCI) is an early indication of progression into Alzheimer disease (AD). Early detection of MCI can help older adults receive appropriate interventions and thus enhance quality of life and the capacity for independent living. Trends in daily routines/activities provide measurement of cognitive and functional status, particularly in older adults. This paper proposes a model to predict the AD progression at a future time point, through leveraging sensor-captured older adult's activity trend data along with health assessment details recorded at various historical time points. The proposed work aims to leverage inherent temporal nature of this disease in predicting the progression. Despite several studies have proposed various approaches for early detection of MCI utilizing activity data, they fail to leverage this temporal nature in predicting disease's future status. Long short-term memory recurrent neural networks (RNN) based prediction model is adopted in the proposed work. RNN models recognize temporal pattern in longitudinal data and hence RNN models are natural choice in predicting AD progression.