Author + information
- Received March 13, 2007
- Revision received June 26, 2007
- Accepted July 2, 2007
- Published online October 23, 2007.
- Petra S. Niemann, MD, PhD⁎,
- Luiz Pinho, MD⁎,
- Thomas Balbach, MD†,1,
- Christian Galuschky, PhD†,1,
- Michael Blankenhagen, PhD†,1,
- Michael Silberbach, MD, FACC⁎,
- Craig Broberg, MD, FACC⁎,
- Michael Jerosch-Herold, PhD⁎ and
- David J. Sahn, MD, MACC⁎,2,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. David J. Sahn, L608, Pediatric Cardiology, Oregon Health & Science University, 3181 SW Sam Jackson Park Road, Portland, Oregon 97239-3098.
Objectives We tested a newly developed 4-dimensional (4D) right ventricular (RV) analysis method for computing RV volumes for both 3-dimensional (3D) ultrasound (US) and magnetic resonance (MR) images.
Background Asymmetry and the anatomical complexity of the RV make accurate determination of RV shape and volume difficult.
Methods Thirty patients, 14 with grossly normal cardiac anatomy and 16 with major congenital heart disease, were studied at the same visit with both 3D echocardiography (echo) and magnetic resonance imaging (MRI) for RV size and function. Ultrasound images were acquired on a Philips 7500 system (Philips Medical Systems, Andover, Massachusetts) with a matrix-array transducer (real-time 3D echo) with full volume sweeps from apical and subcostal views. Sagittal, 4-chamber, and coronal views were derived for contour detection (all 12 to 24 slices). The MR images were acquired with a 3-T MRI magnet with segmented cine-loop gradient echo sequences in short- and rotated long-axis views to cover the RV inflow, body, and outflow tract. The RV volumes were analyzed with the new software applicable to 3D echo MR images.
Results New software aided delineation of the RV free wall, tricuspid valve, RV outflow tract, and apex on 3D echo volumes. Although there was a slightly higher variability measuring right ventricular ejection fraction (RVEF) and volumes obtained by US compared with MRI, both imaging methods showed closely correlated results. The RVEF was measured with 4% variability for US and 5% variability for MRI with a correlation coefficient of r = 0.91. The RV end-diastolic volume was measured at 70.97 ± 15.0 ml with 3D US and at 70.06 ± 14.8 ml with MRI (r = 0.99), end-systolic volume measured 39.8 ± 10.4 ml with 3D US and 39.1 ± 10.2 ml with MRI (r = 0.98).
Conclusions The new RV analysis software allowed validation of the accuracy of 4D echo RV volume data compared with MRI.
Accurate quantification of right ventricular (RV) volume and function has remained clinically challenging despite advances in cardiac imaging. Assessing RV function remains difficult because of the fundamental complex structural geometry of the RV. The RV is composed of several anatomic and functional subunits extending from the tricuspid valve annulus to the proximal os infundibulum (right ventricular outflow tract [RVOT]), then extending from the RVOT to the pulmonary valve as well as the RV body extending to the apex (1–3). Data from embryological observations suggest, however, that the RV infundibulum and RV sinus are distinct areas evolving from distinctly different parts of the embryonic heart (4,5). Electrophysiologic studies indicate that the activation of the RVOT occurs relatively late in systole, possibly being the latest activated area in the heart (6).
Although cardiac magnetic resonance imaging (MRI) is considered the gold standard in the evaluation of RV in the adult population (7,8), 3-dimensional (3D) echocardiography (echo) should be applicable for quantification of RV volumes, stroke volumes (SV), and ejection fraction (EF). Three-dimensional ultrasound (US), as a tool for volumetric assessments of the fetal and adult heart, has shown consistent results to validate cardiac measurements (9,10). More recently, rapid acquisition real-time 3-dimensional echocardiography (RT3DE) has provided the opportunity for functional studies (11).
Our study was designed to evaluate the anatomic shape and function of the RV with a new 3D RV analysis tool evaluating RV volumes and EF, including children with congenital heart disease (CHD) and basically healthy adults as well as adults with coronary heart disease. Our goal was to image the RV with 3-T MRI with a robust multiplanar acquisition protocol and then to use the data as a gold standard for assessing RV volume, EF, and SV by RT3DE.
The patient group consisted of 30 patients with either grossly normal cardiac anatomy (14 patients) or CHD (16 patients) (tetralogy of Fallot, transposition of the great vessels, post-Mustard, atrial septal defect [ASD], ventricular septal defect [VSD]) studied at the same visit with both 3D US and MRI imaging to assess RV size and function (Table 1).The mean age was 39 ± 22 years SD in the adult group and 9 ± 6 years SD in the CHD group. All CHD patients underwent cardiac MRI also for clinical purposes, whereas most of the adult patients were volunteers undergoing cardiac MRI and 3D US for study purposes only. No patient of the adult population was sedated, but all children underwent some degree of conscious sedation, which was administered for the MRI but continued to be effective during 3D US acquisition. The research protocol was approved by the Oregon Health & Science University Institutional Review Board; patients and volunteers enrolled in the study after they signed an informed consent.
All US images were acquired with a Philips 7500 US 3D system (Philips Medical Systems, Andover, Massachusetts) and full matrix-array 3D transducer in a full volume sweep from apical or subcostal views covering the entire RV. The MRI acquisition was performed as a gold standard comparison in a series of cine loops taken at short axis covering the RV from the atrioventricular valve plane to the apex; long-axis views parallel to the interventricular septum in the RV to define the tricuspid valve, and rotated long-axis views to cover the RVOT as well as the tricuspid valve annulus. These specific multiplanar short- and long-axis magnetic resonance (MR) images were acquired to cover the RV body, apex, outflow tract, and tricuspid valve area in all subjects to serve as an anatomic gold standard for RV size and function comparison. Only views covering the complete RV body, RVOT, and tricuspid valve were included for analysis.
All MRI studies were performed with a Philips 3-T (Philips Medical Systems) MR scanner. An electrocardiogram (ECG)-triggered gradient echo cine sequence (repetition time 50 ms, echo time 12 ms, flip angle 60°) was used to acquire all images in contiguous 5- to 8-mm slices from the tricuspid valve plane in short-axis to the RV apex, and frontal views were obtained from the chest wall to the mid left ventricle, allowing a reproducible acquisition of the heart for comparison between patients and for serial studies of the same patient. Also, phase-encoded velocity MRI sequences perpendicular to the RVOT and pulmonary artery were assessed in 17 patients so that MRI RV SV could be further verified by another method that was not volumetric. This multiplanar MRI sequence was used as a gold standard reference for obtaining RV volumes and shapes, by 3D echo (12).
The US was performed in each patient after the MRI. The ECG, respiratory rate, and blood pressure were recorded throughout the imaging process for quantitative evaluation of MRI and echo data, which was undertaken with a prototype TomTec (Unterschleissheim, Germany) 4-dimensional (4D) RV analysis software program at an independent computer work station, which was compatible with Microsoft-based computers. All 3D US data were imported and analyzed in a full volume set; all images were viewed in sagittal (to outline the tricuspid valve in the best possible view), 4-chamber (to outline the apex), and coronal (to outline the RVOT) views, in the best possible view containing a full volume set. The RV volumes were calculated by summing the areas for each slice through the complete volume data set. Each volume data set was imported into the application and manipulated by rotating, angulating, and slicing in any of the 3 displayed orthogonal planes, which would simultaneously reset views in the other 2 planes. A central reference point could be selected by moving through any plane. Secondary rotation was possible. In this way, the clearest possible views were obtained to start the full volume analysis. Contrast, zoom, shading, and color were used to improve delineation of the endocardium. The RV analysis program displayed 3D US as well as MR images in sagittal, 4-chamber, and coronal views for semiautomatic contour detection (Figs. 1 and 2).⇓⇓The software analysis used a semiautomated border detection algorithm with manual correction options based on in vivo normal as well as pathologic RV modeling that was performed to design this program.
The computer-aided analysis for 3D echo combines 3D full volume sweep, US tissue, and the geometric information embedded in a 3D data set to create a surface geometry by 3D. The end-diastolic and end-systolic frames were chosen in all views by visual inspection of cine loops. Although tracing was done in 1 plane, the software reconstructed the traced points in orthogonal planes. Measurements were made on the original full volume echo images and not from a reconstructed model. The endocardial border of the RV was outlined at end-diastole and at end-systole at all levels; 3D images were then constructed with the possibility for manual correction of all frames of the cycle. In all RV 3D images, volume and EF could also be assessed in systole and diastole. Endocardial traces in each slice extended from the interventricular annulus to the ventricular outflow tract. The interventricular septum was excluded from RV volume assessments. Trabeculations were included in the endocardial rim, but the apical component of the RV moderator band was excluded from the cavity.
Measurements for end-diastolic volumes (EDV), end-systolic volumes (ESV), and SV were obtained with the same TomTec program that was adapted to be applicable to MR images. In addition, all RV MRI data were analyzed and compared with a standard Simpson RV MRI disk summation method by measuring the RV free wall area in each short axis slice by manually outlining the epicardial and endocardial boarders. Volumes were calculated by summing the areas for each slice multiplied by slice thickness.
Further phase-encoded velocity MRI data for RV SV as pulmonary artery forward flow was computed at the Philips MRI workstation in 17 patients and compared with the SV measures obtained with 3D US by the TomTec 4D RV method.
The mean ± SD obtained for each parameter in the adult and CHD group with MRI and US were then calculated. The RV volumes and EF measures obtained with US and MRI were compared with each other with the Bland-Altman method. Measurement error and percentage measurement error were determined. A Bland-Altman comparison was also made between the results of 3D US and phase-encoded velocity MRI as well as between 3D US and the standard Simpson summation MRI method for validation purposes.
Interobserver agreement was measured, correlating the results of 2 different researchers for both MRI and 3D US volumes and EF for this novel, anatomically correct RV 3D analysis method, phase-encoded velocity MRI, and the standard Simpson RV summation method.
There was no preselection of cases on the basis of 2-dimensional (2D) image quality; even patients with difficult or suboptimal 2D studies were accepted for 3D imaging and analysis. All MRI data reviewed were acquired in multiplanar ECG gated views as a gold standard to cover the entire RV, showing consistent reproducibility of RV EDV, ESV, SV, and EF data.
Of the 30 studies performed with US and with MRI, all images could be quantitatively analyzed. All images were acquired in a timely fashion with only 20 to 30 min between MRI and US. Reconstructed MRI and echo images were acquired in an image sequence to obtain standardized compatible measures. The US images were acquired in a standardized sequence of images. There was a learning curve for post-acquisition data manipulation, and considerably more time was spent in the volume analysis early in the study period (40 vs. 10 min later), which was attributable to familiarization with software and spatial orientation in 3D echo. All 3D US data and MRI data were analyzed independently from each other.
Visible endocardial boarder definition was essential for the semiautomated boarder detection algorithm to aid RV volume calculation. Difficult-to-visualize RV endocardial boarders were not included, because of poor imaging windows. No intravenous contrast was administered.
Regardless of RV shape or size, visualization of the tricuspid valve area, RVOT, and apex were feasible in all patients. Thirty subjects were analyzed, 16 children with CHD (postoperative tetralogy of Fallot, postoperative transposition of the great arteries, patent ductus arteriosus, VSD, ASD), ages 6 weeks to 14 years, and 14 adults ages 18 to 69 years with normal cardiac physiology or coronary heart disease. The RV EDV and ESV volume, SV, and EF were obtained from each adequate image set. The RV myocardial volume was compared between MRI and US in each patient. The RV free wall, tricuspid valve, RVOT, and apex could be visualized well with both methods, although there was some variability in image quality (Fig. 3).
There was a slight higher variability measuring RVEF and volumes obtained by 3D. The RVEF was measured between 26% and 55% (mean 44 ± 7%) with 3D US and 27% to 55% (44 ± 8%) with MRI. There was a linear correlation between all RV volumes measured with MRI and with 3D US, the correlation coefficient for EDV was 0.99, for ESV 0.98, for SV 0.95, and for RVEF 0.97, respectively. The Bland-Altman method showed only a minimal overestimation of EF by US (Table 2).
The MRI measured EDV, at 70.06 ± 14.8 ml, whereas ESV was measured at 39.17 ± 10.27 ml. The 3D US measured EDV at 70.97 ± 15.06 ml; ESV was 39.88 ± 10.47 ml. Although the RVEF values showed some greater variability with US than with MRI, all volumes measured were closely correlated.
The SV measured from RV volumes was 30.73 ± 8.04 ml with MRI and 31.8 ± 8.56 ml with 3D US. The Bland-Altman method comparing MRI and 3D US acquired SV with phase encoded values showed a close relationship (r = 0.96). These findings are consistent with a close relation between all MRI and 3D US data measured in both groups the adult and pediatric population in regard to SV, EF, and RV EDV and ESV (Fig. 4).
The inter-individual variation seen in SV, EDV, and ESV can be related to the patient’s age and size variation, because this study includes children as well as an adult population. Subdividing these groups into adults (n = 14) and children (n = 16): EDV measured with 3D US ranged from 103.2 to 71.3 ml and with MRI from 105.4 to 74.3 ml in the adults; in children, EDV with 3D US ranged from 67.4 to 43.1 ml and with MRI from 68.3 to 44.2 ml. The ESV measured with 3D US ranged from 62.1 to 35.6 ml in the adults and from 43.2 to 24.6 ml in the children. The ESV measured with MRI ranged from 58.3 to 42.6 ml in the adults and from 44.8 to 20.4 ml in children (Fig. 5).
Phase-encoded MRI velocities of pulmonary artery forward flow obtained in 17 patients were used as an additional standard for SV comparison to validate the results obtained by 3D US and the 4D RV TomTec reconstruction method. Although results for RV SVs still showed a close correlation for all 17 patients (r = 0.96), 4D RV mean ultrasound reconstruction SV was 35.7 ± 6.1 ml, whereas phase-encoded cine velocities SV of pulmonary artery flow were measured slightly higher at 36.9 ± 7.6 ml (Fig. 6).
Although the RVEF showed some greater variability, all volumes measured were closely correlated and reproducible.
Further reference values were obtained in all patients for RV EDV and ESV volumes, RV SV, and RVEF with conventional Philips MRI analysis software and the Simpson’s summation method. We measured the RV volume in each short-axis slice by manually outlining endocardial boarders. Volumes were then calculated by summing the areas for each slice, multiplied by slice thickness. Results showed a close correlation for MRI Simpson’s rule RV EDV and ESV as well as for RVEF and SV in comparison with values obtained by 3D US. The EDV was r = 0.93; ESV, r = 0.92; SV, r = 0.76; but it was worse for RVEF, r = 0.68 (Fig. 7).
The conventional Simpson’s method for volume estimation assumes a prolate ellipsoid shape of the ventricle and has restrictions similar to the M-mode volume area-length method. These mathematical models, although applicable for the left ventricle, are not correct for the RV, which cannot be assumed to be a prolate ellipsoid form but rather represents a form with irregular and variable shape (10,11).
A study comparing 2D and 3D US found that 2D significantly overestimated right heart volume assessments compared with 3D by a factor of 45% and showed a greater reproducibility of 3D (13). The MRI is the gold standard for RV size and especially in CHD patients (14–16). Quantitative 3D echo has been applied for ventricular volume and function assessment in children and adults (17,18); however, no studies have been published comparing 3D US and MRI with measurement tools specifically tailored for the RV. The RVEF measurements in the mid 20% range with both 3D US and MRI are probably attributable to intrinsic disease processes of the RV rather than to the method itself and were clinically supported in 6 subjects.
A close correlation was found between values obtained with 4D RV reconstruction and phase-encoded cine velocity measures of forward pulmonary artery flow. Phase encoded velocities were slightly higher, because early diastolic backflow was reversed in the area between the sampling plane and the pulmonary valve was not subtracted.
To our knowledge this is the first study comparing RV volumes, shape, SV, and EF with a 4D semiautomated contour detection program written specifically for the RV. Our results indicate that 3D/4D echo is reproducible and accurate in measuring RV volumes in healthy adult subjects. Furthermore, we have demonstrated that, 3D/4D echo is capable of also measuring RV volumes, SV, and EF accurately in patients with CHD and complex anatomy and is reproducible.
Although we believe that the anatomical approach we took for 3D echo and multislice MRI works best, we also compared our 3D echo results with standard 5 to 6 plane Simpson’s rule, considered by some to be the “gold standard.” We believe the lower correlation for EF and SV to Simpson’s rule for 3D echo relates to variable inclusion/exclusion of parts of the RVOT and poor anatomic delineation of these structures on transverse only views.
Both intraobserver variability and interobserver variability (3% and <10%, respectively) are reassuring for system accuracy. Volumetric assessments on 3D data can be done fairly rapidly and accurately. A fairly rapid learning curve of the operator was seen affecting accuracy and time. The tendency for some overestimation in the 3D US study group likely indicates an inherent difficulty in excluding trabeculations and the RV moderator band precisely; however, this only resulted in minor variation that was not statistically significant. Furthermore, 3D US is a 3D volume acquisition method, whereas MRI is a high-resolution 2D summation method. Most patients evaluated were children with CHD or adults with normal physiology or coronary heart disease. Patients with very difficult anatomical windows or pulmonary hypertension were excluded in this study to obtain reproducibly compatible results.
Live 3D US is a robust, accurate, and reproducible modality for RV volume and function measurements. Reference values could be obtained for RV volumes and function in children and adults with CHD and normal anatomy, comparing 3D US and MRI, with good correlation between both methods. Also a robust multiplanar MRI protocol was used to validate RV SVs obtained by 4D RV reconstruction of 3D echo versus phase-encoded MRI velocity–based pulmonary artery flow–derived RV SV values.
The method we used should be helpful for RV functional studies both for MRI and 3D echo.
- Abbreviations and Acronyms
- congenital heart disease
- end-diastolic volume
- ejection fraction
- end-systolic volume
- magnetic resonance imaging
- right ventricle/ventricular
- right ventricular outflow tract
- stroke volume
- Received March 13, 2007.
- Revision received June 26, 2007.
- Accepted July 2, 2007.
- American College of Cardiology Foundation
- McAlpine W.A.
- Ho S.Y.,
- Nihoyannopoulos P.
- Foale R.,
- Nihoyannopoulos P.,
- McKenna W.,
- et al.
- Chen P.S.,
- Moser K.M.,
- Dembitsky W.P.,
- et al.
- Mogelvang J.,
- Stubgaard M.,
- Thomsen C.,
- et al.
- Lorenz C.H.,
- Walker E.S.,
- Graham T.P. Jr..,
- et al.
- Bhat A.H.,
- Corbett V.,
- Carpenter N.,
- et al.
- Kuhl H.P.,
- Schreckenberg M.,
- Rulands D.,
- et al.
- Schindera S.T.,
- Mehwald P.S.,
- Sahn D.J.,
- et al.
- Shiota T.,
- Jones M.,
- Chikada M.,
- et al.
- Salustri A.,
- Spitaels S.,
- McGhie J.,
- et al.