NaCTeM

Manchester British Heart Foundation Centre of Research Excellence

Introduction

The Manchester British Heart Foundation Centre of Research Excellence (BHF CRE) will deliver scientific outputs of the highest quality that will benefit global populations with, or at risk of, cardiovascular diseases. Since the mid 2010s, and with substantial BHF support, cardiovascular research in Manchester has been transformed in scope and capability.

The BHF CRE blends long-standing BHF-supported Manchester excellence (for example in genomics; basic discovery science in cardiac pathophysiology) with emerging interdisciplinary strengths (cardiovascular/cerebrovascular-inflammation; cardiovascular data science), building on existing core capacities and partnerships to undertake research spanning from molecules to populations. The CRE includes world-leading computer scientists and engineers as PIs, allowing the incorporation of transformational developments in computational modelling and AI into activities.

The CRE is fully integrated into the powerful Manchester Health Innovation ecosystem, including the Manchester Academic Health Science Centre, that works in close partnership between the NHS, all Manchester universities, and the civic authorities to connect discovery science, industry, and healthcare providers throughout our highly diverse city region of 3 million people.

Themes involving NaCTeM

NacTeM is involved in two of the five themes covered by the CRE, i.e,:

Cardiovascular Data Science

We will leverage large-scale electronic health record (EHR) resources, including powerful Greater Manchester specific datasets, to support interdisciplinary research questions across the CRE. We will utilise state-of-the-art statistical and machine learning approaches in exemplar patient groups to:

  • Develop new methodologies applicable across the spectrum from basic to clinical translational research, underpinning the analytical capacity of other Themes
  • Leverage both national and unique regional Data Science assets and expertise in Manchester, to develop and validate risk prediction models
  • Quantify cardiovascular health inequalities due to geography, social class, race and comorbidity (particularly cancer), map quality of cardiovascular care at the local and national scale, and model the potential impact of new interventions and changes in clinical practice.

Computational Modelling, Simulation and Large Language Models (CMSL)

The CMSL Theme will develop and uniquely integrate state-of-the-art data-driven methods (e.g., signal processing, computer vision and natural language processing), with physics-driven methods (e.g., computational biomechanics and physics-informed learning) to capture information and knowledge. We will focus on exemplar unmet clinical needs, some of which are already identified, and others of which will emerge from research questions across the Centre. We will initially focus on three disease areas: a) congenital heart disease (e.g., aortic coarctation); b) non-congenital structural heart diseases (e.g., valvulopathies and aortic stenosis), c) ischemic and haemorrhagic stroke.

Work in this theme will include:

  • Disease Knowledge Discovery via LLMs for Integrative Analysis and Modelling - We will develop a clinical LLM specific to the cardiovascular domain (CardioLLM). We wil use LLaMA architecture as backbone model; further training will be carried out with (i) text data from PubMed (abstracts), PMC Open Access (scientific articles), EHRs, and UK Biobank; (ii) imaging data from EHRs
  • Deployment of CardioLLM as a foundational model - We deploy will CardioLLM as a foundational model for NLP tasks in the cardiovascular domain such as cardiovascular entity extraction, relation extraction, and summarisation of the scientific literature. NLP tools developed for these tasks will be openly available.
  • Adult congenital heart disease co-pilot - We will apply CardioLLM as an AI co-pilot to support care delivery in adult congenital heart disease

Funding

£8 million (£4 million from the British Heart Foundation and £4 million from the University of Manchesster).

News

Press release from the University of Manchester describing the project.