Premium
TD‐P‐018: MAPPING BEHAVIORAL SYMPTOMS IN DEMENTIA USING PASSIVE RADIO SENSING AND MACHINE LEARNING
Author(s) -
Vahia Ipsit
Publication year - 2018
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2018.06.2034
Subject(s) - computer science , dementia , real time computing , wireless , wireless sensor network , activities of daily living , artificial intelligence , physical medicine and rehabilitation , medicine , telecommunications , computer network , physical therapy , disease , pathology
Background: Impairments of gait and balance often progress through the course of dementia, and are associated with increased risk of falls. Regular assessment of gait and balance could therefore be informative in tracking changes in functional status, and identifying individuals at a high risk of falling to allow for preventative measures. We have developed a technology, called AMBIENT, which enables the frequent, accurate, unobtrusive, and cost-effective measurement of gait and balance parameters. The objective of this study was to demonstrate the feasibility of using AMBIENT for frequent assessment of mobility in people with dementia in a residential facility. Methods: We conducted a pilot longitudinal study with 20 participants (age: 76.9 6 6.7 years, female: 50%) in the geriatric psychiatry unit at the Toronto Rehabilitation Institute, an eighteen-bed inpatient dementia care unit for older adults with behavioral symptoms. The AMBIENT setup included radio frequency identification to identify study participants and a Microsoft Kinect sensor to track body posture. The system automatically monitored participants’ gait as they walked within the view of the sensor during their daily routine and computed the spatiotemporal parameters of gait. Demographic and baseline descriptive measures were collected and falls events tracked. Results: On average, 97 walking sequences per person were collected over a length of stay of 466 37 days. There were 14 falls among study participants: 12 participants did not fall during their length of stay, 4 fell once, 2 fell twice, and 2 fell 3 times. Quantitative measures of gait were stride length (0.8 6 0.1 m), stride time (1.4 6 0.2 s), cadence (89.36 18.1 steps/min), velocity (0.66 0.1 m/s), step length asymmetry (1.2 6 0.6), and step time asymmetry (1.2 6 0.5). Conclusions:This pilot study demonstrates the feasibility of longitudinal tracking of gait over time in a residential dementia setting. Our long-term goal is to translate longitudinal gait parameters onto fall-risk measures. Machine learning techniques will be used to build a robust, multivariate predictive model capable of detecting changes in mobility and falls risk.