Novel Computational Framework for Free-Breathing & Ungated Dynamic MRI

Pulmonary hypertension (PH) is a chronic disease with high mortality. Several pulmonary vasodilators that reduce morbidity and mortality, especially with earlier initiation, have been FDA approved. However, the identification of individuals that would benefit from such therapies currently requires extensive testing of both the heart and the lungs using multiple modalities, resulting in high healthcare costs and delay in diagnosis. This proposal seeks to introduce an MR imaging imaging protocol to diagnose PH within a single imaging session by providing assessments of both cardiac and pulmonary systems. Current MRI methods have several limitations in the above setting. This renewal application aims to overcome these drawbacks using a novel generative SToRM (g-SToRM) framework, which capitalizes on the recent advances in deep generative models and unsupervised learning. This framework significantly improves the analysis manifold regularization framework (SToRM) for cardiac MRI, developed in the previous project. The proposed imaging methods will be validated by comparisons against current breath-held MRI and CT imaging protocols. The preliminary utility of the quantitative metrics to predict PH will also be determined.