Machine Learning–Driven Imaging of Cardiac Microstructure in Diabetes and Heart Failure with Preserved Ejection Fraction

Supervisors:

• Dr Maryam Afzali  mad37@le.ac.uk
• Professor Huiyu Zhou hz143@leicester.ac.uk
• Profesor Gerry McCann  gpm12@leicester.ac.uk

Project description:

Diabetes and cardiometabolic disorders are major contributors to heart failure, particularly heart failure with preserved ejection fraction (HFpEF), which remains poorly understood and challenging to diagnose early (Upadhya & Kitzman, 2020; Shah et al., 2016). Patients with type 2 diabetes (T2D) often develop early myocardial remodelling, including changes in fibre alignment, sheetlet orientation, and extracellular matrix composition, which precede overt structural or functional abnormalities detectable by conventional imaging. Early detection of these microstructural changes may be critical for risk stratification, timely intervention, and personalised management of cardiometabolic patients. Diffusion MRI (Basser & Pierpaoli, 2011) enables non-invasive characterisation of myocardial microstructure by capturing fibre orientation, sheetlet architecture, and tissue anisotropy (Sosnovik et al., 2009; Afzali et al., 2024, 2025), providing insights into early pathological processes associated with diabetes, obesity, and metabolic syndrome.
This project aims to leverage machine learning to detect and predict subtle myocardial microstructural alterations in individuals with T2D and other patient groups with Stage B and mild symptomatic HFpEF. The student will optimise diffusion MRI acquisition and post-processing pipelines to robustly quantify parameters including mean diffusivity (MD), fractional anisotropy (FA), helix angle (HA), and secondary eigenvector angle (E2A) (Nielles-Vallespin et al., 2017; Gotschy et al., 2021). 

Machine learning approaches will be applied to extract latent patterns of microstructural organisation and predict trajectories of myocardial remodelling. In addition, emerging large language model (LLM) approaches will be investigated for integrating multimodal datasets, including imaging-derived features, clinical variables, and unstructured health records. These models have shown promise in automated extraction of cardiac imaging parameters and clinical data interpretation (Wahi et al., 2025), and may enable improved interpretability, automated reporting, and translation of complex imaging findings into clinically actionable insights. By integrating diffusion-derived metrics with functional imaging data such as strain, T1/T2 mapping, and conventional cardiac MRI markers, alongside clinical and biochemical parameters, the student will develop interpretable models linking microstructural changes to cardiac performance. These predictive frameworks aim to identify early disease signatures before clinical symptoms or overt imaging abnormalities appear, supporting proactive patient management and personalised therapeutic strategies.

The project will employ both cross-sectional and longitudinal study designs to capture disease trajectories, evaluating how microstructural alterations evolve over time in relation to glycaemic control, metabolic health, and cardiovascular function. Multimodal datasets will allow correlation of diffusion metrics with structural and functional imaging, blood biomarkers, and exercise physiology measures, providing a comprehensive understanding of early remodelling in diabetes and metabolic disease. By combining high-resolution imaging, AI-based analysis, and advanced statistical approaches, the student will generate robust, reproducible insights into the myocardial microstructure of cardiometabolic patients.
This interdisciplinary PhD provides training across cardiovascular imaging, computational modelling, machine learning, and clinical cardiology. The student will gain hands-on experience in advanced diffusion MRI acquisition and reconstruction, microstructural modelling, AI-driven data analysis, and multiparametric statistical evaluation. Training will include exposure to open-source computational tools, data harmonisation pipelines, and collaboration within Leicester’s cardiovascular imaging group and international research networks. Supervision will be provided by a multidisciplinary team with complementary expertise in imaging physics, AI, and clinical cardiology, ensuring a strong translational focus and alignment with ongoing multicentre initiatives.

The expected outcomes of this project include sensitive, reproducible biomarkers of early myocardial microstructural remodelling in diabetes and cardiometabolic disease, alongside machine learning–based predictive models linking structure to function. These outputs will advance understanding of early disease mechanisms, support personalised risk assessment, and provide a foundation for integrating microstructural metrics into clinical trials and preventative strategies. Ultimately, the project aligns with BHF priorities in translational cardiovascular imaging and aims to bridge the gap between advanced imaging research and real-world clinical application, potentially guiding early intervention strategies for cardiometabolic patients.

References:

1. Upadhya B, Kitzman DW. Heart failure with preserved ejection fraction: New approaches to diagnosis and management. Clin Cardiol. 2020;43(2):145-55.
2. Shah SJ, Kitzman DW, Borlaug BA, et al. Phenotype-specific treatment of HFpEF: A multiorgan roadmap. Circulation. 2016;134:73–90.
3. Sosnovik DE, Wang R, Dai G, Reese TG, Wedeen VJ. Diffusion MR tractography of the heart. J Cardiovasc Magn Reson. 2009;11:47.
4. Basser PJ, Pierpaoli C. Microstructural and physiological features of tissues elucidated by quantitative diffusion-tensor MRI. J Magn Reson. 2011;213:560–570.
5. Nielles-Vallespin S, Khalique Z, Ferreira PF, et al. Assessment of myocardial microstructural dynamics by in vivo DTI. J Am Coll Cardiol. 2017;69:661–676.
6. Gotschy A, von Deuster C, Weber L, et al. CMR diffusion tensor imaging provides novel imaging markers of adverse myocardial remodeling in aortic stenosis. JACC Cardiovasc Imaging. 2021;14:1472–1484.
7. Afzali M, et al. In vivo diffusion MRI of the human heart using a 300 mT/m gradient system. Magn Reson Med. 2024;92:1022–1034.
8. Afzali M, et al. Cardiac diffusion kurtosis imaging in the human heart in vivo using 300 mT/m gradients. Magn Reson Med. 2025.
9. Teh I, Moulin K, Ferreira PF, et al. Multi-centre investigation of cardiac DTI in healthy volunteers by SCMR SIGNET. J Cardiovasc Magn Reson. 2025;101948. 
10. Wahi S, Cross JL, Tysarowski M, Csecs I, Baig M, Ruiz D, Feher A, Kwan JM. A systematic evaluation of open-source large language models for automated extraction of cardiac MRI parameters from unstructured reports. Eur Heart J. 2025;46(Suppl 1):ehaf784.4499.

Please refer to the entry requirements, funding and application advice below before submitting your application