New AI tool targets early dementia detection

A research team is creating an AI powered digital human to help clinicians catch early signs of dementia sooner and more consistently. The system combines structured screening conversations with analysis of facial expressions and physiologic signals to highlight subtle patterns that busy clinics and traditional questionnaires may miss. By turning a routine intake into a standardized assessment, the project seeks to give providers clearer evidence to guide referrals, follow up, and care planning at the first hint of risk.

Researchers are developing AI that can ask screening questions, while observing facial expressions, to evaluate patients for signs of apathy, an early indicator of dementia. Read online

New AI Tool Targets Early Dementia Detection

The effort is supported by the Dementia and Alzheimers Research Initiative at Texas A M Health and led by Mark Benden, a professor of environmental and occupational health at the School of Public Health. The team aims to strengthen one of the largest unmet needs in brain health care, reliable testing at the earliest stages. Rather than replace clinicians, the digital human is intended to extend their reach, bring consistency to observations, and make it practical to screen more people over time with less burden on staff and patients.

Why apathy is a critical early signal

A central focus of the project is apathy, which can emerge before measurable memory decline. Apathy shows up as reduced initiation, lower motivation, and muted emotional engagement. When these changes appear in ways that seem out of character, they can be an early sign of neurodegenerative disease. Catching apathy early creates an opportunity for timely evaluation, lifestyle and care adjustments, and closer monitoring that may slow functional decline or improve quality of life.

Moving from subjective checklists to objective measures

Clinicians already use established questionnaires such as the Apathy Evaluation Scale and the Lille Apathy Rating Scale. These tools provide structure but rely on self report and caregiver impressions, which can be influenced by memory lapses, interpretation of wording, and cultural context. Newer computerized tasks, including the Philadelphia Apathy Computerized Task, offer behavioral measures but still lack validated cut off points for individual level diagnosis. The result is a gap between research insights and actionable findings for day to day practice.

How the digital human works

The project team is building a standardized interview experience that blends conversational AI with multimodal observation. As the virtual interviewer poses questions related to motivation, initiation, and engagement, the system captures objective data streams. These include facial expression dynamics, micro changes in gaze and affect, response timing and hesitation, and selected biometrics collected through approved sensors. The goal is to translate these signals into consistent indicators that can be compared across individuals and tracked for each person over time.

Toward a validated Digital Apathy Signature

At the core of the work is a composite score called a Digital Apathy Signature. It integrates behavioral, cognitive, and emotional dimensions into a single, interpretable risk estimate. By anchoring assessment in observed behavior and physiology rather than only self report, the signature is designed to better separate apathy from look alike conditions, support earlier recognition of dementia risk, and give clinicians a clearer way to monitor change and response to interventions. With validation, this approach could reduce variability, set practical thresholds for action, and help align patients with the right level of follow up.

Clinical value, scale, and workflow fit

A digital human can deliver the same high quality protocol every time, which helps shorten visits, reduce bias, and support triage. In a primary care setting, the tool could flag individuals who might benefit from cognitive testing or specialist referral. In specialty clinics, it could standardize longitudinal follow up, giving teams a straightforward way to see if apathy is worsening, stable, or improving. Because the interview is software driven, sites can scale screenings across larger populations without proportionally increasing staff time.

Guardrails for privacy, equity, and trust

The team is planning for strong privacy protections, transparent model behavior, and human oversight. Data collection will follow informed consent, secure storage, and strict access controls. The assessment content and facial analysis methods will be reviewed for cultural and language fairness to limit bias. Clinicians will remain in the loop for interpretation and decision making, and patients will receive clear explanations of what the tool measures and how results are used.

Collaboration and pathway to validation

Next steps include building the prototype, refining interview content with clinician partners, and conducting studies to validate the Digital Apathy Signature against gold standard scales and clinical outcomes. The team will focus on diverse participant groups to ensure broad reliability. Work will proceed in stages, from feasibility and usability to prospective trials that evaluate predictive value and impact on clinical decisions. The Dementia and Alzheimers Research Initiative at Texas A M Health is committed to advancing research, education, and innovation that can prevent, detect, and treat neurodegenerative diseases. This project reflects that mission by translating laboratory insights into practical tools that fit real world care.

What this could mean for patients and families

For people experiencing subtle changes that are hard to describe, earlier and clearer feedback can ease uncertainty and prompt helpful steps sooner. For caregivers, structured updates on motivation and engagement can inform daily support and planning. For clinicians and health systems, objective measures of apathy can bring consistency to screening, clarify referral thresholds, and enable population level tracking. While this technology will not replace medical judgment, it can supply high quality observations that are often difficult to capture in a brief visit, improving the chances of recognizing risk while there is still time to act.