Clinical measures of cognition typically rely on time-consuming, subjective and expensive assessments. However, our recent advances in computational linguistics, signal processing, and machine learning now allow for objective, automatic, and rapid analysis of cognition, through speech. Our prior work has focused on binary classification problems between people with or without a particular disorder, such as Alzheimer’s disease. In this grant, we will use these modern tools to objectively assess cognition, differentially in people with post-stroke aphasia and memory impairment, by measures of speech and language at 5 time points. This will be applied to a new technological medium – the telephone, which will allow for broader data collection and unique insights in human-computer interaction.
Dr. Frank Rudzicz leads this project, with two industrial partners: WinterLight Labs, and CBI Health Group. WinterLight Labs will supply speech-based assessment to support data collection and data analysis. Patients will be recruited at CBI Health Group clinics, and clinical assessment will be provided by Drs. Regina Jokel and Andrea Iaboni. Research will be conducted at the Toronto Rehabilitation Institute, the Rotman Research Institute at Baycrest, and the University of Toronto.
Our primary objective is to validate the computational speech-based assessment techniques relative to current gold standard assessments. This will involve modern machine learning that is more incisive than the current state-of-the-art. This will be accomplished within two elicitation platforms: a traditional web-based interface, and a phone-based interface. Optimizing the latter is crucial in order to establish the feasibility of using this platform across as wide a population as possible.
In the short term, this project will 1) validate current automatic cognitive assessments, 2) create a new industrial collaboration between a large healthcare network and a local technology startup in Canada, and 3) further our understanding of cognitive function in different patient populations among older Canadians.