Â鶹´«Ã½

Project: Advancing Methods for Health Outcome Assessment with Personalized Humanoids

Challenge

Health Outcome Assessments (HOAs) are standardized psychometric instruments that require in-person interaction between subjects and trained evaluators to assess domains such as mood, activities of daily living, and cognition. While telehealth has enabled digital HOA delivery via tablets, smartphones, and virtual reality, current digital approaches are limited in scope and less reliable than in-person assessments.

How AI Was Used

AI-driven methods are being developed to expand and improve digital HOA reliability, including:

• Fine-tuning automated speech recognition to personalize conversational models using limited individual data
• Designing computer vision models that generate facial representations and track human speakers in real time

Impact

These approaches aim to broaden the scope and improve the reliability of digital HOA batteries for older adults, enabling more accurate, scalable assessments outside traditional clinical settings. 

Project: Next-Generation Digital Detection of Dementia and Mild Cognitive Impairment

Challenge

Dementia and mild cognitive impairment (MCI) remain substantially underrecognized in primary care. In a prospective study, 75% of older adults had previously unrecognized cognitive impairment. Underdiagnosis reduces safety, planning time, and access to appropriate counseling and care.

How AI Was Used

The research team developed a Passive Digital Marker (PDM), an EHR-embedded machine learning algorithm to identify cognitive impairment. While the first-generation PDM was clinically informative, it did not significantly increase documented diagnoses. A second-generation tool, PDM 2.0, integrates structured EHR data with decision-focused summaries extracted from unstructured clinical notes.

Impact

In retrospective validation studies, PDM 2.0 outperformed earlier approaches, achieving AUC values between 0.80 and 0.86. The goal is to increase documentation of MCI and dementia by providing clinicians with patient-specific, actionable insights that support appropriate evaluation and management. 

Let's Stay in Touch.