Why this matters?
Background
Population ageing is a global concern, which brings multiple high‑burden geriatric syndromes that increase public healthcare expenditures and strain societal sustainability.
By 2050, the WHO projects the share of people aged 60+ will nearly double to
22%
in Hong Kong the 65+ group
may reach
35%
Local impact and urgency
A shrinking workforce and narrowing tax base amplify the economic pressure of ageing. Neurocognitive disorders (NCDs) — including age‑related decline, mild cognitive impairment, and dementia — are especially prevalent in older adults. Dementia has an insidious onset and leads to progressive loss of memory, communication and judgment; global care costs (~USD 1 trillion today) are projected to double by 2030.
Limitations of current practice
NCD diagnosis and monitoring rely largely on face‑to‑face neuropsychological testing by clinicians. These assessments are limited by clinician shortages, snapshot measurements that miss intra‑individual variability, reliance on subjective recall, inter‑rater variability, and language/cultural biases.
Our approach
We develop an accessible AI platform that uses spoken‑language biomarkers to screen and monitor cognition remotely. Non‑intrusive speech recordings are analysed by deep‑learning models to detect subtle changes in timing, fluency and coherence — enabling timely alerts and data for clinicians.
How it works
Innovative Evaluation Platform
We will develop an automated, objective, highly accessible evaluation platform using inexpensive, easily acquired biomarkers for NCD screening and monitoring. Remote platform access enables continual monitoring and timely patient alerts between clinic visits.
Longitudinal data and benefits
Collecting individualized longitudinal “big data” allows detection of subtle cognitive changes over time. Early flagging helps prevent under‑diagnosis, improves disease management, delays institutionalization, and can lower care costs.
Spoken‑language biomarkers
NCDs often manifest as communicative impairments, so we target spoken‑language biomarkers as non‑intrusive alternatives to blood tests and brain scans. Speech can be captured remotely; millisecond‑resolution speech event records (latencies, dysfluencies, etc.) provide sensitive measures of cognitive function.
AI methods and novelty
We will build AI‑driven technologies to automatically extract and select multidimensional spoken‑language features — from micro‑hesitations to dialog coherence — using fit‑for‑purpose deep learning. Our approach emphasizes comprehensive dimensional coverage and adaptability across environments to ensure consistent, objective assessments.
This research will provide unprecedented data and technological support for earlier NCD diagnosis and more timely clinical care, aligning with WHO priorities to make dementia a public health and social care focus. Our goal is to reduce the burden of NCD through AI‑enabled tools that better support patients and caregivers in Hong Kong.
Get In Touch
Reach out to us today to learn more about our research initiatives and how our innovative evaluation platform can contribute to the advancement of cognitive health assessment and management.