Affiliation: Biomedical Engineering, OHSU
Funding Period: 2011 - 2013
Funding Source: Roybal Pilot
View related topics
Relative to healthy elders, people with mild cognitive impairment (MCI)—a syndrome that frequently precedes Alzheimer’s Disease (AD)—display impaired lexical access and favor simpler syntactic construction in their speech and writing. Further, epidemiological studies have found that mean differences in prodromal linguistic behavior differentiate between groups who ultimately do or do not develop symptomatic AD decades before diagnosis. Together, these observations suggest that linguistic features associated with AD-pathology are present long before most people are diagnosed through standard practice—consistent with the long time-course of the disease’s neuropathology. There is therefore reason to predict that automatic, continuous monitoring of linguistic behavior has the potential to serve as a first line of action in detecting onset of decline.
As a relatively naturalistic form of written language that is gaining popularity among elders—and one that may be inexpensively analyzed—email is an attractive candidate medium for precisely this sort of longitudinal monitoring. Algorithms to detect signatures of MCI can be seamlessly integrated into email clients on elders’ home computers, rendering them unobtrusive. Our research objective is to build and test a software package that will extract useful information about cognitive health of email users by monitoring the linguistic features embedded within email messages, while safeguarding the privacy of email. Our aims are: 1) Build a de-identified, annotated corpus of outgoing email messages from a two-year timespan for a set of 20 participants recruited from a monitoring study of the oldest old; 2) Conduct a feasibility study of automatic parsing and analysis of emails from participants into a longitudinal dataset; and 3) Conduct and leverage results of preliminary statistical modeling of trajectories over time in normal and MCI-participants toward identifying stable metrics and patterns of longitudinal change that differ between groups.