Clinical Manifestations.

Area of Science:

• Neuroscience
• Computational Linguistics
• Gerontology

Background:

• Aging populations face rising rates of neurocognitive disorders, including subjective cognitive decline (SCD), mild cognitive impairment (MCI), and Alzheimer's disease dementia (ADD).
• Traditional diagnostic methods (cognitive tests, neuroimaging) are costly, time-consuming, and inaccessible in many regions.
• Automated speech and language analysis (ASLA) presents a cost-effective, scalable alternative for early detection in underserved communities.

Purpose of the Study:

• To investigate the efficacy of ASLA in predicting cognitive and neuroimaging measures across the neurocognitive decline continuum in Chilean Spanish-speaking individuals.
• To address the gap in ASLA research for this specific demographic and cognitive spectrum.

Main Methods:

• 150 Chilean participants (17 controls, 55 SCD, 57 MCI, 21 ADD) completed fluency tasks, cognitive assessments (ACE-III, MoCA, IFS), and MRI scans.
• Machine learning models analyzed speech features (word properties, timing) from fluency tasks using the TELL app.
• Models predicted cognitive test scores and brain volumes (gray matter, white matter, hippocampus).

Main Results:

• Significant correlations were found between ASLA-derived features and cognitive scores (ACE-III: r=0.55, MoCA: r=0.39, IFS: r=0.31).
• ASLA also correlated with normalized gray matter volume (r=0.34).
• These findings were most robust using word-property features from combined fluency tasks.

Conclusions:

• A fully automated ASLA pipeline effectively captures cognitive and MRI indicators of brain health from brief speech tasks.
• This approach is examiner-independent, objective, time-efficient, and affordable, offering advantages over standard diagnostic tools.
• ASLA represents a scalable solution to promote global health equity in combating dementia and cognitive decline.