In Vitro Volume Validation

Ten water-filled latex balloons of different size and shapes were used to validate the accuracy of Simpson’s method for 2D echocardiography ( Figure 1 A), 4D Auto LVQ for 3DE ( Figure 1 B), and the Extended MR Workspace software for CMR ( Figure 1 C). To determine the balloon’s volume with Simpson’s method and 3DE, we used the visible inner layer of the latex membrane as the boundary. On CMR images, the real inner layer was not detectable, so the outermost margin of the balloon was used as the boundary. Apparently internal balloon borders seen on CMR images depict an artifact known as Gibbs ringing, which represents a mathematical limitation of the Fourier transform.

Long axis ( upper row ) and short-axis ( lower row ) view of one example latex balloon on 2D echocardiography (A) , reconstructed 3DE (B) , and CMR imaging (C) .

Balloon volumes were assessed on the same day with all methods and then compared with their real volumes. Water-filled balloon volumes were determined by measuring their weights on a laboratory precision balance and subtracting the weights of the empty balloons, assuming that under standard conditions 1 mL equals 1 g pure water weight. The Auto-EF tool could not be tested as described because it needs a small dynamic volume change of ≥10% to work at all.

Study Population

All patients undergoing clinically indicated CMR examinations in our cardiology department within a time span of 1 month were invited to participate in this study ( n = 183). Forty-seven patients agreed to undergo additional comprehensive echocardiographic examinations. Clinical characteristics are shown in Table 1 . The patient population was heterogeneous: 15 patients had normal myocardial function without critical coronary artery disease; furthermore, 14 patients with cardiomyopathy (10 dilated, two hypertrophic, two cardiac amyloidosis), 10 patients with coronary artery disease (two with one-vessel, two with two-vessel, and six with three-vessel disease), four patients with different forms of known cardiac arrhythmia and/or conduction disorder (Wolff-Parkinson-White syndrome, premature atrial contractions, paroxysmal atrial fibrillation, premature ventricular beats), two patients after heart transplantation, one patient with an atrial septal defect, and one patient with systemic vasculitis were included. All patients were in sinus rhythm during both examinations. To minimize time-dependent variation of hemodynamic characteristics, echocardiography was commenced <30 min after CMR.

Standard as well as two new methods for the assessment of LV volumes and function were conducted in the same cohort. Patients with suboptimal echocardiographic image quality were not excluded, to reflect a real-life patient population. The study was carried out prospectively after approval by the Ethics Committee of the University of Heidelberg and in concordance with the Declaration of Helsinki. Written informed consent was obtained from all persons.

Echocardiography

All echocardiographic examinations were performed on a commercially available ultrasound machine (Vivid E9 BT 11; GE Vingmed Ultrasound AS, Horten, Norway) according to the guidelines of the American Society of Echocardiography and using a 1.5- to 4.6-MHz phased-array probe (M5S-D) for 2D imaging and an active-matrix four-dimensional volume phased-array probe (4V-D) for 3DE. Image acquisition was conducted in a breath-hold manner, and at least three consecutive heart cycles were recorded after passive end-expiration. The LV endocardium was used as the boundary for volumetric measurements. Papillary muscles and visible trabeculae were part of the blood pool. If endocardial border detection was uncertain, nonvisible parts were interpolated manually. Three different techniques were used to determine LV volumes and function:

• 1.
  

  The modified biplane Simpson’s method, as recommended by the American Society of Echocardiography : End-diastolic and end-systolic endocardial borders were traced manually on frozen 2D images obtained from the apical two- and four-chamber views to derive end-diastolic volume (EDV) and end-systolic volume (ESV). The LV EF was calculated according to the formula EF = (EDV − ESV)/EDV × 100%.

  
  
• 2.
  

  The Auto-EF tool (EchoPAC version 110.1.1 BT 11; GE Vingmed Ultrasound AS) : Using dynamic 2D images of the apical four- and two-chamber views, three anchor points were set within the LV cavity, two at the level of the mitral valve annulus and one at the LV apex. Endocardial borders were then detected and traced automatically by the software during a whole heart cycle to calculate EDV, ESV, and EF. When needed, corrections could be carried out manually ( Figure 2 A).

  
  

  
  

  
  

  

  
  

  
  

  Endocardial border detection by Auto-EF (A) , 4D Auto LVQ (B) , and manual border tracing of a midventricular CMR short-axis slice (C) in end-diastole ( upper row ) and end-systole ( lower row ). All data sets are derived from the same patient, with a moderately reduced EF.

  
  
  
  
• 3.
  

  Three-dimensional echocardiographic analysis using the 4D Auto LVQ tool (EchoPAC version 110.1.1 BT 11; GE Vingmed Ultrasound AS, Trondheim, Norway) : Three-dimensional data sets were acquired, comprising the whole heart in a 90° × 90° pyramidal scan volume and were recorded during breath-hold in passive end-expiration. A multibeat mode over up to six heart cycles was used to achieve high temporal resolution, with frame rates between 30 and 50 frames/sec. After data acquisition, fully automated orientation of the apical four-chamber, two-chamber, and long-axis planes was attempted. Orientation was manually adjusted if needed. Two points were then manually set in end-diastole as well as in end-systole, one at the level of the mitral valve, one at the LV apex, to define basal and apical limits of the LV cavity. Endocardial borders were detected, and values of EDV, ESV, and EF then calculated automatically by the software. Automatic endocardial border detection was manually corrected if the software excluded papillary muscles or visible trabeculae from the LV cavity ( Figure 2 B).

CMR Imaging

A whole-body CMR scanner (1.5-T Achieva; Philips Medical Systems, Best, The Netherlands) was used for image acquisition, applying a short-axis multislice cine steady-state free precession sequence with parallel imaging (8–12 slices; slice thickness, 8 mm; gap, 2 mm; ≥35 phases per cardiac cycle; balanced fast field echo; repetition time, 2.9 msec; echo time, 1.45 msec; reconstructed voxel size, 1.5 × 1.5 × 8 mm acquisition; sensitivity encoding factor, 2). According to recent guidelines, the LV cavity, defined as the border between compacted and noncompacted myocardium, was manually traced in each short-axis slice in end-diastole and end-systole ( Figure 2 C) with a commercially available software package to calculate EDV, ESV, and EF (Extended MR Workspace version 2.6.3.4; Philips Medical Systems). Papillary muscles were part of the blood pool.

Statistical Analysis

Statistic analysis was carried out with the SigmaStat version 3.5 (Systat Software, Inc, Chicago, IL), and graphs were drawn with GraphPad Prism version 5 (GraphPad Software, Inc, La Jolla, CA). To test for normal distribution, the Kolmogorov-Smirnov test was used. Significance was tested using Student’s t test or the Mann-Whitney-Wilcoxon test in case of non-normality. P values < .05 were considered statistically significant. Method comparison was conducted with linear regression analyses and Bland-Altman plots. Pearson’s correlation coefficient was used for normal distributed values and Spearman’s correlation coefficient in case of non-normality. Twenty randomly chosen data sets were reanalyzed in a blinded fashion by the same and by a second experienced sonographer to assess intra- and interobserver variability. These values are expressed as coefficients of variation and were calculated by the standard deviation of the differences between the two measurements, divided by their mean value and expressed as percentages.