Developing effective interventions for prevention and treatment of Alzheimer's disease (AD) requires early detection of the disease. With recent advances in wearable device and physiological data analytical tools, it is feasible to assess many physiological functions unobtrusively by monitoring spontaneous motor activity. The goal of this project is to develop an integrated, non-invasive biomarker for the risk of Alzheimer's dementia using motor activity recordings. Among many physiological functions derived from motor activity, reduced physical activity levels, sleep disturbances, circadian dysfunction, and perturbation in fractal physiological regulation appear to precede the cognitive symptoms of Alzheimer's disease (AD), and signify an elevated risk of developing Alzheimer's dementia. However, it is unknown whether these dysfunctions predict Alzheimer's risk independently or they are interconnected to amplify/diminish each other's adverse effect. For a better prediction of Alzheimer's dementia using motor activity, PI and his team propose to leverage the above physiological risk factors using a novel artificial intelligence technique. To achieve this, PI and his team will utilize the existing longitudinal database of the Memory and Aging Project at Rush Alzheimer's Disease Center, in which over 1,400 old participants have been enrolled since 2005 and have agreed to (i) undergo annual motor activity monitor and structured clinical examinations and (ii) donate brain, the entire spinal cord, and selected nerve and muscles at the time of death. The ambulatory motor activity recordings collected annually will be used to assess a series of constructs including (i) physical activity (level of physical activity, intensity of physical activity, and average daily inactivity duration), (ii) sleep characteristics (total sleep duration, sleep efficiency, and sleep fragmentation), (iii) circadian rhythmicitiy (normalized 24-h amplitude, acrophase of daily activity rhythm, interdaily stability, and intradaily variability), and (iv) fractal motor regulation (temporal correlations in motor activity fluctuations at small and large time scales). Using these physiological measures together with clinical diagnosis, cognition, genetics, and post-mortem histopathology, three aims will be addressed: 1) determine whether a deep learning based neural network model can construct an integrated biomarker from the above physiological measures for better prediction of the risk of Alzheimer's dementia and the risk of conversion from mild cognitive impairment to Alzheimer's dementia in a short time frame (i.e., 2 years); 2) determine whether the integrated biomarker modifies or interacts with the genetic effect on AD; and 3) determine how specifically the integrated biomarker reflects AD pathology at autopsy. Achieving the aims will result in the first integrated biomarker of motor activity that leverages multimodal, noninvasive measurements for a better prediction of Alzheimer's dementia. The results to be obtained may also lead to a better understanding of the complex biology and physiology of AD, which will potentially guide the seeking of disease modifying therapies or interventions.
Many non-cognitive physiological changes including reduced physical activity, sleep and circadian disturbances, and degraded fractal regulation predict the risk of Alzheimer's disease (AD), and they can be assessed unobtrusively from daily spontaneous motor activity. Adopting a novel artificial intelligence approach, this project is proposed to develop an integrated biomarker for AD risk in older adults by combining these physiological functions, and to examine the relationships of this biomarker with genetics and AD pathology. This project will potentially provide a non-invasive tool for predicting AD risk and lead unique insights into the complex pathology of AD, which is important for developing new therapeutic strategies to treat AD or slow down AD progression.
