Incorporating Voice Analytics Into Cognitive Assessments: Alvaro Pascual-Leone, MD, PhD

WATCH TIME: 4 minutes

"If you say, it’s mild cognitive impairment vs amnestic mild cognitive impairment, that makes a difference between an 80% risk of progression to dementia vs 10% or 30% risk. That’s a huge difference. If we can identify whether or not [it is amnestic]—in addition to signs of frailty—as well as characterize the whole person better, it allows us to be more fine-grained not only from a prognostic point of view, but for guiding interventions.”

For years, the Alzheimer disease community has attempted to find the most highly accurate combination of tools to predict clinical progression and trajectory of cognitive decline. Linus Health’s DCTclock is an FDA registered Class II medical device that assesses an individual’s cognitive health by analyzing more than 700 features of the clock drawing process using artificial intelligence and machine learning. Linus subsequently developed the Digital Clock and Recall Assessment, which incorporates voice and delayed verbal recall alongside the DCTclock.

At the 2022 Alzheimer’s Association International Conference (AAIC), July 31 to August 4, in San Diego, California, a large-scale, 495-patient, longitudinal trial evaluated the added benefit of the DCR and the relationship between voice analytics and diagnosis. Results showed that DCR was more highly correlated with Mini-Mental State Examination scores than DCTclock (r = 0.43 vs 0.38), but also, that combining acoustic features with cognitive assessments achieved greater classification accuracy for healthy (area under the curve [AUC], 0.95), mild cognitively impaired (AUC, 0.93), and Alzheimer disease (AUC, 0.97) groups than models including either acoustic features or cognitive assessments alone.¹

Senior investigator Alvaro Pascual-Leone, MD, PhD, believes this type of approach has serious clinical implications, considering most of the measures used are standard and widely available in the field. Pascual-Leone, who is chief medical officer and cofounder of Linus Health, provided additional background on application of acoustic and speech production to help classify cognitive status, as well as why these results serve as recommendations for future decision-making.

Click here for more coverage of AAIC 2022.

REFERENCE
1. Banks R, Jannati A, Gomes-Osman J, et al. Enhanced machine learning classification of mild cognitive impairment with multimodal digital biomarkers. Presented at: 2022 Alzheimer’s Association International Conference; July 31-Aug 4; San Diego, CA. 60485