Cardiac MRI Evaluation

All MRI examinations were performed using a 1.5-T Siemens Sonata scanner (Siemens Healthcare, Erlangen, Germany) and a five-element cardiac array for signal reception, as previously described. All image acquisitions were electrocardiographically gated and acquired during breath holds. End-diastolic volume, end-systolic volume (ESV), and EF measurements were derived from manual segmentation of short-axis cine images at end-systole and end-diastole (Argus; Siemens Healthcare), and EF was derived using Simpson’s rule. Additionally, the endocardial tracings in conjunction with apical and basal locations from two-chamber, three-chamber, and four-chamber long-axis views were used to generate an endocardial surface of the left ventricle at end-systole and end-diastole. Peak systolic global longitudinal strain (GLS) was calculated using the average of the change in length of the surface from end-diastole to end-systole divided by the lengths at end-diastole.

Echocardiographic Evaluation

After the acquisition of noncontrast echocardiographic images, 0.1 mL Definity (Lantheus Medical Imaging, North Billerica, MA) in slow bolus injections, followed by slow saline flush, was administered for optimal endocardial definition using an iE33 scanner (Philips Medical Systems, Andover, MA). We used standard settings with low to intermediate mechanical index (mean values ranging from 0.3 to 0.4) for contrast recordings, with images optimized to achieve dense and homogenous LV opacification and dark myocardium. The maximum frame rate to achieve LV opacification was used for this study (39 frames/sec). Although higher frames rates would be advantageous for speckle imaging, they cannot be realized with the current scanners and would probably impair LV opacification and endocardial definition. Offline LV volume and EF analysis for noncontrast and contrast images was performed using standard manual tracing of end-systolic and end-diastolic endocardial borders from apical two-chamber and four-chamber images (Xcelera; Philips Medical Systems), and results were obtained using Simpson’s biplane method. In addition, echocardiographic recordings were uploaded to the new software, the two-dimensional (2D) cardiac performance analysis (CPA) for speckle-tracking imaging (frame rate, 39 frames/sec; TomTec, Munich, Germany). Two-dimensional CPA has been developed for noncontrast echocardiography. This technique works very similar to Velocity Vector Imaging (Siemens Healthcare). Two-dimensional CPA tracks speckles mainly at the endocardium. This is different from speckle-tracking as implemented in many clinical scanners, and it also tracks speckles in the myocardium. The operator scrolled through the cardiac cycle and, irrespective of cardiac cycle, identified a single frame with good endocardial definition, which was usually during systole. A minimum of 10 border points, at the operator’s discretion, were depicted for manual tracking. Then tracing of the endocardial border throughout the cardiac cycle was automatically performed, and LV volumes and EF were computed. In addition, speckles were tracked in a frame-by-frame fashion, and traces depicting segmental longitudinal strain were computed automatically. After the automatic tracking throughout the cardiac cycle, recordings were reviewed to identify studies with inappropriate tracking. No corrections of the tracked contour were made. The tracking was regarded as unsuccessful if the tracked contour on end-diastolic or end-systolic frames was >5 mm away from the visually perceived border of the LV cavity in two or more of six segments in the four-chamber or two-chamber view.

Systolic GLS was obtained by averaging the segmental strain curves at maximum instantaneous peak. Similarly, noncontrast images from the same patients were also analyzed with the software and strain parameters were recorded.

LV volumes and EF derived from 2D CPA were compared with those obtained using MRI and the contrast images analyzed with Xcelera using Simpson’s biplane rule. Mean GLS from contrast echocardiographic recordings were compared with those obtained from noncontrast echocardiographic images. Two experienced physicians assessed the image quality for noncontrast recordings, with optimal quality defined as adequate visualization of all myocardial segments. In a subset analysis for GLS, only data from patients with optimal image quality on noncontrast echocardiographic recordings were included for comparison between contrast and noncontrast measurements. In another subset of 22 patients, in which MRI analysis had also been performed, GLS strain data from contrast and noncontrast echocardiographic recordings were correlated with those derived from MRI.

Reproducibility Analysis

All 2D CPA measurements were repeated in a subset of 20 patients by a second physician and by the same primary reader, who were blinded to data. Repeated measurements were performed ≥2 weeks after the first evaluation. As mentioned previously, the selection of the cardiac cycle for the manual border detection was left to the operator’s discretion. Interobserver and intraobserver variability was calculated as the absolute difference of the corresponding pair of repeated measurements and as a percentage of their means in each patient and then averaged over the total number of patients, and correlation was tested using intraclass correlation coefficients.

Data are presented as mean ± SD or percentages as appropriate. Student’s t tests were used for comparisons of continuous data. Linear regression analysis was used to correlate measurements among different techniques, with agreement assessed using Bland-Altman analysis and intraclass correlation coefficients, as described previously.

All measurements were analyzed using SPSS version 18 for Windows (SPSS, Inc, Chicago, IL). P values < .05 were considered statistically significant.