Eye-tracking AI device aids in detection of dementia risk in underrepresented group

Key takeaways:

• Population-based study examined 29 individuals, eight of whom met criteria for mild cognitive impairment.
• Eye-tracking metrics outperformed standard testing when education level was removed as a covariant.

A novel, eye-tracking metrics device aided in the detection of dementia risk in a small population of Colombian individuals with differing levels of education, according to a poster presented at CTAD.

“A key factor precluding the validity and generalizability of novel developments in the field of early diagnosis is their cultural bias,” Mario A. Parra, MD, PhD, neuroscientific officer at ViewMind, and colleagues wrote.

Parra and colleagues tested the efficacy of the ViewMind AI Solution, a novel vision-based device that utilizes neurocognitive biomarkers, artificial intelligence and eye-tracking metrics to assess dementia risk, among a cohort of individuals in Colombia.

The study included 29 individuals, eight of whom met criteria for mild cognitive impairment, as well as healthy controls matched according to age but who differed in education level.

All participants underwent clinical, functional and neuropsychological assessments, in addition to ViewMind, which recorded eye-tracking metrics during memory binding exercises involving color and aided in assessment of low-level integrative memory ability.

Researchers used multivariate analyses to determine group differences in neuropsychological, cognitive and biomarker (eye-tracking metric) domains, with education as a confounder.

According to results, researchers found significant group effects in the noncontrolled model of the neuropsychological domain [F = 2.69; P = 0.034; beta = 83%], which did not occur when education was included as a covariant [F = 1.78; P = 0.143; beta = 61%].

However, group differences in the noncontrolled model were significant in the biomarker domain [F = 3.83; P = 0.007; beta = 94%] and increased after removing education as a covariant [F = 4.10; P = 0.005; beta = 95%]. These findings were evident only during the encoding phase of the integrative memory task.

“This profile mirrors that seen in patients with or at risk of Alzheimer’s disease dementia,” Parra and colleagues wrote. “These results support our proposal of ViewMind AI Solution as a culture-free neurocognitive biomarker for the early detection of dementia in underrepresented populations.”