FLAME (Fair, Lightweight, Affordable, Multimodal, and Explainable) Machine Learning Methods for Predicting Alzheimer’s Disease Using Personalized and Adaptive Multimodal Biomarkers
Description: <p><strong>Content and objectives of the research</strong></p><p>The project’s main goal is to develop FLAME: a Fair, Lightweight, Affordable, Multimodal, and Explainable machine learning framework for predicting AD. FLAME addresses several current challenges, including limited diversity in existing individual datasets, high infrastructure requirements, the complexity of integrating biological data, high diagnostic costs, and poor model transparency. To ensure and verify the FLAME’s generalisability, reproducibility, and efficacy, we apply it to up to 14 datasets, covering a broad range of ethnic groups, ages, and disease stages. The system is designed to work even when only partial or low-cost data are available, and to provide clear explanations about which factors contribute most to disease risk or progression. FLAME will also help clinicians determine which additional tests, if any, may be useful in supporting a diagnosis. Finally, we aim to integrate FLAME into clinical practice.</p><p> </p><p><strong>Scientific and societal context</strong></p><p>Early detection and personalised clinical support are essential for AD management, as no curative AD treatment currently exists. This project aims to design an AI tool that is accessible, transparent, and usable across different healthcare environments, including those with limited resources. It seeks to support global efforts to improve AD diagnosis and care in an equitable and responsible way.</p>
Amount: CHF 491,800
Funding type: research
Funder scheme: Project funding
Start date: 2026-01-01
End date: 2029-12-31
Investigator: Chén, Oliver Y.
Institution: University of Lausanne – LA
Funded outputs count: 0
Funder award ID: 10006845
OpenAlex ID: G6562063209