Deep learning-based synthesis of hyperpolarized gas MRI ventilation from 3D multi-inflation proton MRI
Joshua R Astley1,2, Alberto M Biancardi2, Helen Marshall2, Malina-Maria Tofan1, Laurie J Smith2, Paul JC Hughes2, Guilhem J Collier2, Matthew Q Hatton1,3, François-Xavier Blé4, Rod Hughes5, Jim M Wild2,6, and Bilal A Tahir1,2
1Oncology and Metabolism, University of Sheffield, Sheffield, United Kingdom, 2POLARIS, Sheffield, United Kingdom, 3Clinical Oncology, Weston Park Cancer Centre, Sheffield, United Kingdom, 4Translational Science and Experimental Medicine, Research and Early Development, Respiratory & Immunology, BioPharmaceuticals R&D, AstraZeneca, Cambridge, United Kingdom, 5Clinical Development, Research and Early Development, Respiratory & Immunology, AstraZeneca, Cambridge, United Kingdom, 6Insigneo Institute for in silico medicine, University of Sheffield, Sheffield, United Kingdom
Hyperpolarized gas MRI can visualize and quantify regional lung ventilation with exquisite detail, but clinical uptake is limited to a few centres worldwide. Alternative, non-contrast techniques have been proposed to image ventilation, including 1H-based surrogates of ventilation from multi-inflation 1H-MRI. Recently, deep learning has shown potential for generating synthetic images in multiple modalities within the lung image analysis field. We propose a 3D multi-channel deep learning approach to synthesize hyperpolarized gas MRI and assess its quantitative performance using several common image synthesis metrics across a large, diverse dataset of lung pathologies using 5-fold cross-validation.
This abstract and the presentation materials are available to members only;
a login is required.