Feasibility of automated voice based cognitive assessment on a consumer voice platform
In clinical studies, accuracy and consistency in scoring and administration of assessments are still problematic. Computerised cognitive assessment batteries such as CANTAB have demonstrated the utility of automating test administration and scoring.
Minimising human errors and automating the scoring and administration of cognitive tests can help to improve the detection of patients in the prodromal phase of Alzheimer’s disease, enable high-frequency monitoring of changes in cognitive function and improve the statistical power of clinical trials.
Methods
For this research, presented at AAIC 2017, we developed a short voice-based cognitive battery with tasks assessing episodic memory and working memory using the Amazon Alexa voice platform. The tasks were based on traditional neuropsychological assessments.
We assessed 20 participants aged between 20 and 73 using the automated battery and simultaneous independent audio recording. Three participants were discarded due to audio recording failure, leaving an analysis sample size of 17.
The participants were also assessed using a standard CANTAB battery measuring similar cognitive domains (Paired Associates Learning, Spatial Working Memory and Rapid Visual Processing).
Results
The results showed an excellent correspondence between automated and manual scoring of participant performance – showing that automated voice-based assessment of cognition may be feasible for population-level studies and home-based monitoring.
Interested in learning more about the feasibility of automated, voice-based cognitive assessment on a consumer voice platform? Download this poster.
