CVD-RAG: An AI Prototype for Cardiovascular Disease Clinical Decision Support Integrating Retrieval-Augmented Generation with Large Language Models

Overview

Cardiovascular disease (CVD) is Australia's leading cause of death, accounting for over 25% of all mortality (AIHW, 2023). Clinicians managing complex CVD cases, particularly patients with concurrent type II diabetes, must synthesise evidence from a rapidly growing body of clinical literature. General-purpose Large Language Models (LLMs) such as GPT-4 and Llama-3 offer compelling natural-language interfaces to this knowledge, but are known to produce hallucinated, ungrounded, or uncited answers, making them unsafe for unsupervised clinical use.

Retrieval-Augmented Generation (RAG) directly addresses this limitation by coupling an LLM with a live document retrieval system: before generating a response, the system retrieves the most relevant passages from a curated knowledge base and grounds the answer in those sources, with traceable citations. RAG has rapidly become the dominant architecture for safe, knowledge-grounded LLM deployment in high-stakes domains.

This project aims to design and develop a RAG-based AI prototype that enables users to pose CVD-related clinical questions and receive evidence-based, citation-supported responses. It will demonstrate how modern natural language processing (NLP) and machine learning techniques can be effectively applied to address real-world challenges in healthcare.

Research engagement

Literature review, data collection and a system demo of the AI prototype.

Research activities

Survey & Design (survey the RAG methods for medical NLP literatures and design a RAG-based model)

Data & Implementation (assembles CVD corpus from PubMed Central (PMC) open-access articles via the NCBI API (freely available), supplemented by university library sources where accessible).

Demo & Reporting (deliver a minimal end-to-end working demo, conducts an informal qualitative evaluation and writes the final report documenting the system design, findings, limitations, and future research directions).

Research skills

Students will gain hands-on experience with advanced AI methodologies, including RAG, systematic model evaluation, and AI-driven system design. They will develop practical skills in integrating LLMs with external knowledge sources, constructing retrieval pipelines, and implementing mechanisms for evidence grounding and citation.

Outcomes

This project aims to develop a RAG-based AI prototype that answers CVD-related clinical questions with evidence-based, citation-supported responses, demonstrating the application of modern NLP and machine learning to real-world healthcare challenges.

Skills and experience

Applicants should have a strong academic background in computer science or a related field. Prior experience with programming (e.g., Python) and familiarity with machine learning or natural language processing concepts is highly desirable.

Start date

2 November, 2026

End date

19 February, 2027

Location

QUT GP, S Block

Additional information

The students will meet with supervsiors weekly  face to face or online.

Keywords

• Machine learning
• LLM-based Information retrieval
• LLM-based Information extraction
• Decison support

Contact

Professor  Yuefeng Li,  School of Computer Science

07 31385212

y2.li@qut.edu.au