We propose to develop an AI-powered question answering system that enables personalized health monitoring, prediction, and treatment recommendations. Our research will leverage the rich biomedical and lifestyle data in the UK Biobank to answer individual-level health queries in natural language, integrating predictive models for disease risk, progression, and treatment outcomes. Research questions include:
RQ1: How can large-scale cohort data be used to personalize health predictions?
RQ2: What machine learning approaches best translate population-level insights to individual recommendations?
RQ3: How can question answering systems deliver accurate, explainable, and clinically relevant answers to users?
Objectives:
1. Build models that predict disease risk and treatment response using UK Biobank data.
2. Develop a question answering interface that retrieves personalized insights grounded in these models.
3. Ensure transparency, explainability, and alignment with clinical guidelines.
4. Validate the system’s performance against established risk prediction benchmarks.
Scientific rationale:
Existing risk calculators and clinical decision tools are limited in personalization and accessibility. By applying advanced machine learning and natural language processing to the UK Biobank dataset, we aim to make health predictions more precise and usable. The project will demonstrate how large-scale biobank data can directly support individualized preventive medicine and treatment planning, improving health outcomes and patient engagement.
