Image analyses
All MRI examinations were transferred to an offline workstation with the commercial postprocessing software, CVI42 (Circle Cardiovascular Imaging Inc., Calgary, Canada). Endocardial and epicardial contours rendered by automated analyses on short-axis cine images at end-diastolic and end-systolic phases were manually reviewed and corrected as necessary. The trabeculae and papillary muscles were included in the ventricular blood pool. The most basal short-axis slice measurement was defined as having an LV circumference of at least 270° surrounded by myocardium. The most apical short-axis slice measurement was defined as the last slice where the LV cavity was visible. LV morphologic and functional parameters, including the ejection fraction (EF), end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), and myocardial mass, were obtained. All parameters were indexed to body surface area (BSA).
LGE was considered present if seen in both axial planes and extended beyond localized ventricular insertion areas. The patterns were classified as linear, patchy, or a combination of the two. The locations were classified as ventricular septal, LV free wall, or as occurring at both locations. The distribution was classified as midmyocardium and non-midmyocardium. The mean signal intensity and standard deviation (SD) were derived, and a threshold of \(\geq\)5SD above the mean was used to define as the LGE areas. Adding the LGE areas of all short-axis slices yielded the total volume (g), which was also expressed as a proportion of the total LV myocardium (% LGE).
All CMR analyses were performed by two independent radiologists (with 15 and 5 years of experience in cardiovascular radiology, respectively), and a third radiologist provided adjudication, if necessary. All CMR analyses were performed by observers blinded to the clinical and 24-h ECG data.