Project Summary In the United States, atrial fibrillation (AF) is the second most common cardiovascular condition in adults, after hypertension, affecting 33 million individuals worldwide, with more than 7 million patients in the U.S. and Europe, over 400,000 U.S. hospital admissions per year, 5 million office visits per year, 276,000 emergency room visits per year, a $26 billion annual increase in U.S. healthcare costs, and a five-fold and an almost two- fold increase in the risk of stroke and mortality, respectively. A key marker in the disease process is the collagen and fibrosis formation in the left atrium (LA). MR imaging methods have been developed to assess the left atrium, including fibrosis but are still unreliable and controversial. Specifically, late gadolinium enhancement (LGE) MRI has played a key role as a non-invasive imaging tool for in vivo visualization and quantification of atrial fibrosis. However, the inconsistent image quality and the significant amount of expert- level supervision involved in the fibrosis quantification process are major roadblocks to its reproducibility. This proposal offers a new MRI imaging technique and an associated machine learning approach to better assess the left atrium and to determine the repeatability of these measurements. Aim 1 will introduce robust and repeatable MRI acquisition and reconstruction methods for obtaining isotropic resolution in a fixed time of LGE images of the left atrium. Unique motion compensation models will be integrated into the reconstruction methods for the LGE data. Aim 2 will provide accurate machine learning based methods for segmenting the left atrium wall and assessing atrial myocardium fibrosis. Aim 3 will rigorously test and further refine the non- invasive fibrosis assessment methods in an experimental canine model with fibrosis. Aim 4 will quantify the reproducibility of the new acquisition, reconstruction, and post-processing methods compared to existing methods in patients. This proposal will directly impact clinical science and standards of care in cardiology and AF management through significant improvements to LGE imaging of left atrial fibrosis. Technological contributions of this work will further impact the field of biomedical image analysis through the improvement of MRI-based techniques for imaging cardiac tissue structure. To promote transparency and reproducibility, the proposed efforts will be released as open-source tools consistent with the principles of reproducible research and open science practices. This open-source nature will further make this project a catalyst for future methodological innovations and clinical investigations. Our long-term goal is to streamline and standardize atrial fibrosis quantification for the clinical management of AF patients, and this project will establish the groundwork for achieving this goal.