The chief medical officer and cofounder of Linus Health provided background on the reasoning for the measures used in a new machine learning algorithm that classifies cognitive status. [WATCH TIME: 3 minutes]
WATCH TIME: 3 minutes
"We thought it would be paramount that the assessment is short, works well in the workflow of primary care physicians, and frankly, that the assessment is something that they are familiar with and has good backing in the literature.”
Previously conducted studies have used acoustic and speech production as a means to classify cognitive status, but results have been inconsistent. At the 2022 Alzheimer’s Association International Conference (AAIC), held July 31 to August 4, in San Diego, California, one study examined if the addition of speech production metrics can increase the classification accuracy of cognitive impairment. The group of investigators, led by Alvaro Pascual-Leone, MD, PhD, created an enhanced machine learning algorithm using several multimodal digital biomarkers that categorized patients as either cognitively healthy (HC)-cognitively impaired (CI), HC-probable Alzheimer disease (AD), HC-mild cognitive impairment (MCI), and MCI-AD.
At the conclusion of the analysis, statistical models that included both digital clock drawing metrics and 3-word speech production analysis to classify cognitive status outperformed models that included only digital clock drawing metrics. The study included several cognitive assessments and metrics, including the Rey Auditory Verbal Learning Test, Delayed Recall, Mini-Mental State Exam, physician diagnosis, Functional Activities Questionnaire, Digital Clock Test, and Digital Clock and Recall.
Pascual-Leone, chief medical officer and co-founder of Linus Health, sat down with NeurologyLive® to answer questions on the reasons behind the metrics chosen. He also discussed the need to find the right combination of assessments and how this translates to clinical care.