PROJECT SUMMARY: The overall goal of this application is to develop, implement and test a “single button push”, integrated combination of innovative MRI solutions to enable widespread and generalizable implementation of quantitative evaluation of chronic liver disease in < 5 minutes. We aim to design a reliable, efficient, low variability, and fully automated, MRI exam. This goal will be enabled by artificial intelligence (AI), reengineered chemical shift encoded (CSE)-MRI to provide “error-free” free-breathing measurement of liver fat and iron, an innovative MRI suite design, and automated analysis. In this way, we aim to achieve high-throughput, low-cost evaluation of liver disease with high accuracy, precision and reproducibility. Abnormal accumulation of triglycerides in hepatocytes, or steatosis, is the earliest feature of non-alcoholic fatty liver disease (NAFLD), affecting ~100 million people in the US. Liver iron overload is common in patients with hereditary hemochromatosis and those receiving repeated blood transfusions. Early, affordable, and accessible non-invasive detection and quantitative staging of liver fat and iron would impact the health of millions of people at risk for NAFLD and its comorbidities, as well as those with liver iron overload. Confounder-corrected CSE-MRI provides simultaneous estimation of liver proton density fat fraction (PDFF) and R2*, which are accurate, precise and reproducible biomarkers of liver fat and iron. A primary determinant of the cost of MRI is scheduled MRI suite time. Minimum slot times to accommodate the majority of patients are driven by variability in exam duration and MRI suite turnaround time. As MRI scan times are shortened, the largest contributor to exam duration is the time needed for i) manual image prescription, ii) repeated scans (rework), and iii) room turnaround time. Many patients, including children, are unable to hold their breath for the duration of CSE-MRI (~20 seconds) leading to ghosting artifacts that corrupt PDFF / R2* maps, necessitating repeated CSE-MRI acquisitions and exacerbating exam time variability. We will address these challenges by developing fully automated AI-based image prescription based on multi-center, multi-vendor data at 1.5T and 3T, in parallel with a novel “error-proof” high SNR “snapshot” CSE-MRI method that is insensitive to breathing motion. This will be performed using a novel MR “Smart Suite” design, capable of patient turnaround in less than 2 minutes, followed by automated quantitative analysis and reporting. We will implement and test a fully automated, single button push CSE-MRI exam by aiming to: 1). Develop and optimize motion insensitive, high SNR, free-breathing CSE-MRI for accurate and precise measurement of PDFF and R2*, 2). Confirm the accuracy, repeatability, and reproducibility of the proposed CSE-MRI method in patients with liver fat and iron overload, and 3). Implement and validate a fully automated CSE-MRI protocol in less than 5 minutes of MR room time. If successful, this work will provide a high-throughput, high value solution for liver fat/iron quantification. The innovations proposed in this application will also have broad applicability beyond CSE-MRI, and ultimately reduce cost and increase access, through improvements in MRI scanner utilization.

PROJECT NARRATIVE: This application will develop and integrate a series of technical solutions that improve efficiency and reduce variability of current magnetic resonance imaging (MRI) exams, into a true 5-minute clinical exam. We will train artificial intelligence models to enable single button-push MRI exams, using a multi-center, multi-vendor approach to ensure the reproducibility and generalizability of our results. We focus on achieving these goals on the development of a rapid and quantitative MRI method that can detect and quantify early features of fatty liver disease and liver iron overload, to facilitate preventative treatment for millions of Americans who would otherwise go undiagnosed.