MRC DTP in Precision Medicine

LLMs-Driven Early Detection of Dementia: Leveraging Multi-source Language Data from Spontaneous Language to EHR-based Longitudinal Clinical Narratives

Supervisors

Prof Honghan Wu, School of Health and Wellbeing, University of Glasgow
Prof David McAllister, School of Health and Wellbeing, University of Glasgow
Prof Terence Quinn, School of Cardiovascular and Metabolic Health, University of Glasgow

Summary

Dementia is a progressive condition that affects memory, thinking, and everyday activities, with Alzheimer’s disease being the most common form. More than 50 million people worldwide live with dementia today, and this number is expected to triple by 2050. The impact extends beyond the individual, placing huge emotional, social, and financial strain on families, healthcare systems, and societies. While there is currently no cure, detecting dementia early can make a big difference. Even delaying the onset by just five years could reduce the number of cases by up to 40%, easing pressure on patients, carers, and health services.

Today’s diagnostic tests are often expensive, invasive, and impractical for large-scale use. However, language provides a promising new way forward. Subtle changes in speech and communication often appear before other symptoms, making language a powerful early marker of dementia. This project will use the latest advances in artificial intelligence (AI), particularly large language models, to analyse two key sources of language: patients’ everyday speech and doctors’ notes in health records. By combining these perspectives, the research aims to develop practical, non-invasive, and cost-effective tools that help detect dementia earlier, support better care planning, and ultimately improve quality of life for patients and their families.