Dr Qiang Zhang
Oxford BHF CRE Basic Science Transition Fellow
- Start Date: 01/07/2021
- End Date: 30/09/2023
- BHF CRE Mentors: Stefan Piechnik, Vanessa Ferreira, Stefan Neubauer & Thomas Nichols
Research project title: Towards Replacing Late Gadolinium Enhancement with Virtual LGE for Gadolinium-Free CMR Tissue Characterisation
Research summary:
I am a deep learning scientist with cross-domain knowledge of cardiovascular diseases and CMR physics. My fellowship project aims to develop artificial intelligence (AI) imaging techniques to replace traditional late gadolinium enhancement (LGE) for myocardial tissue characterisation.
LGE on cardiovascular magnetic resonance (CMR) imaging is used routinely to characterize myocardial tissue and guide the treatment of heart diseases. However, LGE requires intravenous injection of a gadolinium contrast agent, which prolongs the scan, increases the cost, and is cautioned in some patients.
Using advanced AI techniques, we are developing a “virtual native enhancement” (VNE) imaging technique to replace traditional LGE for gadolinium-free myocardial tissue characterization. The new technique combines MR images that do not normally need contrast injections, and uses AI to train machines to predict what an LGE image would look like. First results on hypertrophic cardiomyopathy patient data show that the approach can produce as clear or even better-quality images than traditional LGE, providing clinicians with the same information. We are further collecting large and representative CMR datasets and developing deep learning models to address a range of cardiac pathologies, such as myocardial infarction.
AI has been a main driving force for many clinical breakthroughs. Development of AI technologies for cardiovascular research requires close collaboration between technicians and clinicians. The University of Oxford Centre for Clinical Magnetic Resonance Research (OCMR) has cross-disciplinary teams of magnetic resonance specialists, cardiologists and AI scientists to support the development and validation of VNE techniques.
Future Prospects:
Beyond this fellowship, I plan to scale up the project to address more myocardial pathologies. I also aim to expand the concept of AI-driven “virtual contrast dye” to replace more CMR imaging techniques that currently rely on contrast agents. Implementation of the prototype on MRI scanners could pave the way for wider VNE protocol dissemination and multi-centre clinical trials. With further development, this may lead to the next generation of gadolinium-free CMR scan protocols that are faster, cheaper, and more patient-friendly.
Undergraduate Degrees/Training:
PhD, MEng, BEng, BSc
Recent publications
-
Gadolinium-free Virtual Native Enhancement for chronic myocardial infarction assessment: independent blinded validation and reproducibility between two centres
Conference paper
THOMPSON P. et al, (2023), Global CMR 2024 Scientific Sessions
-
Quality control-driven framework for reliable automated segmentation of cardiac magnetic resonance LGE and VNE images
Conference paper
Gonzales RA. et al, (2023)
-
TVnet: a deep-learning approach for enhanced right ventricular function analysis through tricuspid valve motion tracking
Conference paper
Gonzales RA. et al, (2023)
-
Quality control-driven deep ensemble for accountable automated segmentation of cardiac magnetic resonance LGE and VNE images
Journal article
Gonzales RA. et al, (2023), Frontiers in Cardiovascular Medicine
-
Myocardial Strain Measurements Derived From MR Feature-Tracking: Influence of Sex, Age, Field Strength, and Vendor.
Journal article
Yang W. et al, (2023), JACC Cardiovasc Imaging