| The public health implications of growing numbers of older adults at risk for dementia place pressure on identifying dementia at its earliest stages so as to develop proactive management plans. However, this capability of early detection is very challenging with the contemporary detection process in the form of conventional clinician visits, which are typically not sensitive to detecting mild cognitive or functional decline. Fortunately, some studies have documented that changes in motor capabilities precede and may be indicative of cognitive impairment. A major motivation for this thesis was: can we utilize changes in motor capabilities and activity patterns to automatically predict the onset of dementia in older adults through continuous monitoring using unobtrusive sensors?;Many studies have demonstrated that the prodromal dementia phase commonly identified as mild cognitive impairment is an important target for this early detection of impending dementia amenable to treatment. Consequently, in this thesis we explored different approaches for the automatic detection of mild cognitive impairment using unobtrusive sensing technologies. We started off by exploring the feasibility of detecting mild cognitive impairment in older adults using a number of predefined measures associated with their in-home walking speed. We were able to achieve this goal with an area under the ROC curve and an area under the precision-recall curve of 0.97 and 0.93, respectively, using a time frame of 24 weeks.;We were also very interested in exploring the feasibility of detecting mild cognitive impairment in older adults using changes in their activity patterns. We used inhomogeneous Poisson processes to build generalized linear models of older adults' activity that would model their presence in different rooms throughout the day. Using these models, we extracted very interesting insights and visualized significant changes in older adults' activity patterns as they started experiencing cognitive impairment. Finally, we devised a method for automatic detection of mild cognitive impairment using changes in older adults' activity models, and we achieved this goal with an F 0.5 score of 85.6 percent. We also investigated the two subtypes of mild cognitive impairment, namely the amnestic and the non-amnestic subtypes, and made very interesting observations. |
