Automated Short-term Longitudinal Analyses of Speech and Language in Dementia
What's the challenge?
Major symptoms of Alzheimer’s disease (AD) and other dementia are displayed in an individual’s speech. Specifically, patients with mild cognitive impairment (MCI) (i.e. preclinical stage of dementia) and early-stage AD show subtle, speech-based symptoms like word-finding difficulties. Unfortunately, current gold standard cognitive assessments, like the Mini Mental State Exam (MMSE), are not sensitive enough for detecting the subtle changes in speech that are suggestive of cognitive impairment.
What have we done?
Winterlight Labs (WLL) has created a speech-based cognitive assessment app that can detect mild changes in speech. This app will be tested as a cognitive assessment for seniors diagnosed with MCI, AD, or other dementia. This study investigated whether changes in cognition over time relate to changes in language over time, and whether changes in different aspects of language can be reported to families and caregivers in a way that helps them make decisions at home for seniors.
What did we find?
Monthly cognitive assessments using the app show that, on average, residents’ language, memory, and thinking abilities remained fairly stable over the course of the 7-month study. The predicted MMSE scores for residents obtained from their Winterlight assessment follow the same trend as their actual MMSE scores. This study suggests that Winterlight’s assessment is easy to complete with persons with dementia, reliable, more engaging, and faster than gold standards. In addition, the algorithm can correctly differentiate between residents with AD and dementia, using their speech profiles.