Data-Driven Cardiovascular Flow Modelling: Echocardiography and AI for Precision Treatment of Congenital Heart Disease

Background

Congenital heart disease (CHD) is the most common birth defect, affecting more than 1.3 million newborns each year worldwide. While survival rates in high-income countries have improved dramatically, CHD still causes over 200,000 deaths annually in low- and middle-income countries (LMICs), where up to 90% of patients lack timely access to clinical care. Even in well-resourced healthcare systems, clinicians face challenges in determining which patients require urgent surgical intervention and how to optimise treatment strategies. 

Clinical diagnosis and monitoring of CHD relies heavily on echocardiography, an inexpensive and widely available imaging modality that provides real-time data on heart structure and blood flow. However, echocardiography cannot capture complete flow information in the aortic arch, a region critically important in conditions such as aortic coarctation and interrupted aortic arch. Missing or incomplete data reduces clinicians’ ability to stratify patients by disease severity, delaying or complicating surgical decisionmaking. 

Advanced methods such as computational fluid dynamics (CFD) and 4D MRI can provide complete haemodynamic data but are too resource-intensive for routine clinical use, particularly in LMICs. Recent advances in machine learning, particularly neural operator approaches, offer a transformative opportunity: by learning from CFD-derived datasets, algorithms can rapidly infer complete blood flow patterns from incomplete and noisy echocardiographic measurements. This enables generation of patient-specific, real-time flow reconstructions at the point of care. 

This project will build a precision medicine tool that integrates echocardiographic data, patient-specific vascular geometries, and machine learning models trained on CFD simulations. By providing clinicians with comprehensive, interpretable flow information, the tool will support accurate patient stratification, improve surgical planning, and reduce reliance on more invasive or costly imaging. 

Aims

The overarching aim is to develop and validate an AI-augmented flow modelling tool for congenital heart disease that enables precision diagnosis and treatment planning. Specific objectives are: 

1. Identify key vascular and fluidic metrics that determine flow characteristics in the aortic arch, using CFD simulations and existing datasets. 
2. Develop and train a neural operator model capable of predicting 3D flow fields in the aortic arch from incomplete echocardiographic data. 
3. Validate the tool against CFD simulations for idealised and patient-specific aortic geometries. 
4. Integrate the approach with real patient data, initially from CT-derived geometries and echocardiographic measurements, with future extensions to ultrasound-derived geometries. 
5. Assess robustness to noisy or variable input data and evaluate the ability of the model to generalise across diverse patient anatomies. 
6. Demonstrate clinical utility by applying the tool to CHD cases such as aortic coarctation and interrupted aortic arch, focusing on its potential to inform timing and type of surgical intervention. 

By aligning advanced computational modelling with clinical imaging and AI, this project will create a new pathway for precision cardiology, directly addressing the DTP’s themes of interdisciplinarity, methodological innovation, and tackling major chronic diseases and health inequalities. 

Training Outcomes 

This project is inherently interdisciplinary, offering a unique training environment across engineering, computer science, and clinical medicine. The student will gain: 

• Quantitative and computational skills: expertise in CFD simulations, data-driven modelling, neural operator architectures, and integration of physics-informed AI. 
• Biomedical domain knowledge: understanding of cardiovascular physiology, echocardiographic imaging, and clinical challenges in CHD management. 
• Precision medicine expertise: experience in developing patient-specific models for disease stratification, linking computational outputs to clinical decision-making. 
• Responsible AI training: skills in data governance, interpretability, and maintaining clinician oversight in AI-augmented tools. 
• Collaborative and translational research experience: opportunities to work closely with clinicians at partner hospitals, computational scientists at the Edinburgh Parallel Computing Centre, and international collaborators, preparing them for leadership roles in academia, healthcare, or industry. 

By the end of the PhD, the student will be equipped to lead interdisciplinary projects at the interface of AI, computational modelling, and healthcare, with a strong focus on precision medicine applications. The tool developed in this project has the potential not only to improve outcomes for children with CHD but also to establish a transferable framework for other cardiovascular and ultrasound-based conditions, maximising impact on global health and health equity.