Project ID BE-MI2026_09

ThemeBE-MI

Co Supervisor 1A Prof Oleg Aslanidi Faculty of Life Sciences & Medicine, School of Biomedical Engineering & Imaging Sciences, Digital Twins for HealthcareEmail

Co Supervisor 1B Dr Adelaide de Vecchi Faculty of Life Sciences & Medicine, School of Biomedical Engineering & Imaging Sciences, Digital Twins for HealthcareEmail

Predictive Modelling of Thrombus Formation Using Physics-Informed Neural Networks

Background: Thromboembolism and stroke are a leading cause of morbidity and mortality, accounting for over 30% of all deaths in recent decades. Mechanisms of such diseases depend on multiple inter-related factors, often connected to cardiac arrhythmias such as atrial fibrillation. Biophysical computer models have been developed to understand these mechanisms and predict risks but their complexity means they are generally incompatible with the clinical timescale and impractical to use in the clinic. Statistical deep learning (DL) models have been used as an efficient way to overcome the drawbacks of traditional mechanistic models and numerical methods. In particular, Physics-Informed Neural Networks (PINNs) can combine precise formalisms of biophysical equations with computational efficiency of DL.

Aims: This project’s overarching goal is develop PINNs as a novel DL tool for simulating thrombus formation in the cardiovascular system. The project aims to: 1) develop a PINN framework that combine advantages of the mechanistic and statistical models to simulate blood coagulation risks efficiently, 2) train and test the PINN using in-silico patient cohorts, and ultimately 3) evaluate these novel PINNs on patients data and predict the associated risks of stroke.

Timeline/Objectives: Year 1: Rotational MRes project to develop DL neural networks for emulation of 2D atrial models (see below). Year 2: Development of Physics-informed neural networks (PINN) for fast simulation of 3D atrial blood flow. Year 3: PINN application to simulate patient-specific blood dynamics and thrombogenesis risks. Year 4: Evaluation of the PINN predictions of stroke risks in a retrospective patient cohort, thesis write-up.

Skills: The candidate will learn coding skills in Python and Matlab, image processing skills, and gain a detail understanding of different DL architectures. They will also achieve in-depth knowledge of the physiological mechanisms of blood coagulation and thrombus formation. Finally, they will understand how new technology can be designed to be compatible with, and address, clinical needs.

MRes Project: This will prepare the student for the following PhD project by applying DL neural networks trained on patient medical imaged and metrics from image-based 2D atrial models to predict the patient outcomes directly from the images, thus emulating the time-consuming models.

Representative Publications

[1] Muffoletto M, Qureshi A, Zeidan A, Muizniece L, Fu X, Zhao J, Roy A, Bates PA, Aslanidi O. (2021). Toward Patient-Specific Prediction of Ablation Strategies for Atrial Fibrillation Using Deep Learning. Frontiers in Physiology 12: 674106; DOI: 10.3389/fphys.2021.674106

[2] Nazarov I, Olakorede I, Qureshi A, Ogbomo-Harmitt S, Aslanidi O. (2022) Physics-informed Fully Connected and Recurrent Neural Networks for Cardiac Electrophysiology Modelling. 2022 Computing in Cardiology (CinC) 498: 1-4; DOI: 10.22489/CinC.2022.188

[3] Ogbomo-Harmitt S, Muffoletto M, Zeidan A, Qureshi A, King AP, Aslanidi O (2023) Exploring interpretability in deep learning prediction of successful ablation therapy for atrial fibrillation. Frontiers in Physiology 14: 1054401; DOI: 10.3389/fphys.2023.1054401

[4] Qureshi A, Lip GYH, Nordsletten DA, Williams SE, Aslanidi O, de Vecchi A. (2023) Imaging and biophysical modelling of thrombogenic mechanisms in atrial fibrillation and stroke. Frontiers in Cardiovascular Medicine 9: 1074562; DOI: 10.3389/fcvm.2022.1074562

[5] Qureshi A, Melidoro P, Balmus M, Lip GYH, Nordsletten DA, Williams SE, Aslanidi O, de Vecchi A. (2025) MRI-based modelling of left atrial flow and coagulation to predict risk of thrombogenesis in atrial fibrillation. Medical Image Analysis 101: 103475; DOI: 10.1016/j.media.2025.103475

[6] Melidoro P, Sultan ARA, Qureshi A, Yacoub MH, Elkhodary KL, GYH Lip, Aslanidi O, de Vecchi A. (2024) Enhancing stroke risk stratification in atrial fibrillation through non‐Newtonian blood modelling and Gaussian process emulation. The Journal of Physiology 287283: 1-19; DOI: 10.1113/JP287283