Pediatric Cardiopulmonary MRI using RF Navigators and High Dimensional Deep Learning

ABSTRACT The cardiac and pulmonary systems are inherently linked through the pulmonary vascular system which leads to secondary pulmonary disease in cases of cardiac pathology. This is especially the case in congenital heart disease where the pulmonary blood supply is often substantially altered by abnormal outflow tracts and

ventricular formation. Cross-sectional imaging has proven to be invaluable for assessing pediatric diseases, including congenital heart disease; however, current cardiopulmonary evaluations typically require multiple exams (SPECT, echo, MRI, and CT) to evaluate the cardiovascular and pulmonary systems. Each exam adds

risk to already fragile patients, introduces complex logistics of performing multiple exams, and can delay care of patients who may require urgent management. Often, only a subset of exams are performed, and clinical management is based on incomplete information and disregards the strong potential for cardiopulmonary

coupling. In this project, we aim to develop MRI methods that can simultaneously and efficiently evaluate both anatomy and function in pediatric cardiopulmonary diseases. MRI is theoretically well suited for quantitatively imaging both the cardiac and respiratory systems but is traditionally challenged by its slow imaging speed and

sensitivity to artifacts. Recently, our group has proposed methods for dramatically more robust lung imaging using the combination of ultrashort echo time MRI with advanced motion corrected reconstruction strategies. In this proposal, we extend these techniques and introduce novel methods to provide improved and comprehensive

diagnostics of the entire cardiopulmonary system. First, we introduce a free-running approach to cardiopulmonary imaging to provide anatomical imaging and the quantifications of ventilation, perfusion, cardiac function, and respiratory resolved cardiac flow dynamics. We specifically aim to image continuously with T1

weighted and velocity encoded sequences, and subsequently reconstruct this data with a high-dimensional deep learning approach. The reconstructions use novel motion corrected methods to directly estimate images and apply deep learning in a highly compressed space. Secondly, we aim to develop next-generation motion

management using an RF navigator technique, Beat Pilot Tone, that can be applied during any pulse sequence to measure bulk, respiratory and cardiac motion. Beat Pilot Tone provides a basis for motion tracking that enables improved imaging efficiency, a simplified setup without cardiac leads or respiratory belts, and much better

measures of bulk motion. These techniques will be evaluated in normal control participants and pediatric subjects with congenital heart disease, each with comparisons to state-of-the-art imaging. The impact of this project is to shift the paradigm for clinical management of cardiopulmonary diseases to a single-scan comprehensive imaging

study and supporting an integrated assessment of interaction between the pulmonary and cardiac systems in disease. While this is targeted at pediatric cardiopulmonary diseases, the innovations can be applied broadly to MRI studies throughout age ranges and to other studies that suffer from motion artifacts throughout the body.