Calculation of left ventricular (LV) ejection fraction (EF) by Doppler stroke volume and end-diastolic volume (EDV) derived from LV diastolic diameter (LVIDD) could be reliable and feasible in clinical practice. In subjects with a wide range of LV volumes and EFs, magnetic resonance imaging (MRI) was used to evaluate the accuracy of common formulas (Teichholz and Z method) to estimate EDV from LVIDD (EDV Teich ; EDV Z ) versus volumetric EDV. The accuracy of simplified Doppler-EF was tested in a separate study sample versus real-time 3-dimensional (RT3D) echocardiography and versus bidimensional Simpson’s method. A new equation to calculate EDV from LVIDD was derived using MRI and tested in the RT3D echo samples. Feasibility of Doppler-EF was tested in a third sample of consecutive inpatients and outpatients. In the MRI sample, EDV Teich was greater whereas EDV Z was smaller than volumetric EDV (both p <0.01); however, a quadratic equation estimated EDV from LVIDD with accuracy (R 2 = 0.97). In the echocardiographic sample, independent of severe segmental wall motion abnormalities, EDV Teich was greater whereas EDV using Simpson’s method was smaller than RT3D EDV (all p <0.05); Doppler-EF using EDV Teich was lower compared with EF by Simpson’s rule or by RT3D-EF (all p <0.01). However, RT3D-EF showed no differences compared with Doppler-EF when EDV was calculated by the novel MRI-derived equation. Feasibility was 95% for Doppler-EF and 72% using Simpson’s method. In conclusion, equations estimating EDV from LVIDD affect the accuracy of simplified Doppler-EF. However, Doppler-based EF may be accurate and feasible even in the presence of LV contractile asynergy.
Left ventricular (LV) ejection fraction (EF) is the most commonly used index of LV systolic function in clinical practice. In clinical studies, LV EF computed from LV linear dimensions has been accepted in the presence of normal LV geometry and in absence of LV asynergy. In contrast, in LVs with asynergy, LVEF should be calculated from LV volumes obtained by planimetry. However, LV volume reconstruction by planimetry is time-consuming, requires optimal endocardial borders visualization, can be heavily affected by foreshortened apical views, and may be affected by suboptimal reproducibility. Alternatively, real-time 3-dimensional (RT3D) echocardiography was found to be more reliable for assessment of EF, but it is a modality not widely diffused and feasible in daily clinical practice. Doppler echocardiography allows the calculation of LV stroke volume (SV) without being affected by ventricular asynergy and dyssynchrony. As proposed by Dumesnil et al, Doppler SV may be used to calculate LV EF (Doppler-EF) in combination with LV end-diastolic volume (EDV), calculated from LV internal diastolic diameter (LVIDD) using the Teichholz’s formula. However, it was demonstrated previously that estimation of EDV from LVIDD using the Teichholz’s equation may also be inaccurate in subjects without LV asynergy. Finally, whether EDV may be accurately estimated from LVIDD in subjects with LV asynergy and abnormal diastolic geometry remains unclear. Accordingly, the aims of the present study were to (1) test the accuracy of EDV estimation from LVIDD by different equations in a learning sample using cardiac magnetic resonance imaging (MRI) as the gold standard, (2) test the impact of EDV calculation from LVIDD on the accuracy of Doppler-EF estimation in a testing sample using RT3D echocardiography, and (3) evaluate the feasibility of Doppler-EF versus EF obtained by using the biplane Simpson’s rule in a separate series of consecutive patients undergoing clinical echocardiography.
Methods
We conducted our study in 3 samples. Study participants were recruited in a clinical setting among inpatients of a cardiology ward or outpatients of a cardiology unit. The participants in the 3 study samples, the cardiac MRI, the echo data set, and the feasibility samples were not recruited simultaneously.
The first step of our study focused on the assessment of the accuracy of formulas calculating EDV from LVIDD. For this purpose, inpatients and outpatients were recruited consecutively among those who underwent cardiac MRI for clinical indications. We aimed to include the widest range of LV volumes, diameters, and EF. Thus, we recruited consecutively with the target of including at least 5 cases with EF ≤35% and at least 5 cases with EF ≥70%; additionally, we aimed to recruit at least 5 cases with EDV ≤100 ml and at least 5 with EDV ≥200 ml; those targets were reached with 39 consecutive cases recruited in approximately 2 months.
A separate study sample, including inpatients and outpatients and healthy volunteers who underwent 2-dimensional (2D) and RT3D echocardiography provided the data set to test the impact of various methods of EDV calculation on the accuracy of Doppler-EF compared with RT3D-EF. The echocardiography sample comprised 41 patients who underwent RT3D, Doppler, and 2D echocardiography. Mitral regurgitation with vena contracta ≥5 mm was an exclusion criterion. Patients with aortic stenosis of any degree or hypertrophic cardiomyopathy were also excluded because aortic anulus or pulsed Doppler spectrum time-velocity interval at aortic anulus could not be used to assess Doppler SV according to the reference method used in the present study.
The feasibility of Doppler-EF was compared with that of LV planimetry using the biplane modified Simpson’s rule to calculate LV volumes in a third sample of 100 consecutive subjects referred for standard echocardiography in our clinical laboratory (40% inpatients, 60% outpatients). All studies were evaluated by a single reader (VP). Feasibility was defined as the ratio between the number of studies with EF evaluated with sufficient confidence compared with the total number of studies. For LV planimetry, a study of sufficient quality was required to define the confidence of estimation of LV EF as sufficient. The quality of LV planimetry was defined sufficient when the endocardial border was well seen in end-systolic and end-diastolic frames in at least 2 of the 3 segments of each LV wall in apical 4- and 2-chamber views. For Doppler-SV, a study of sufficient quality was characterized by an angle of approximately 90° in apical views between the aortic anulus and the Doppler cursor beam, with tolerance predefined at ≤20°; no adjustment of the velocity scale by angle steering procedure was allowed; the Doppler spectrum was required to be clear-cut and with a peak velocity clearly identified; the assessment of aortic anulus required that the aortic valve leaflets were well visualized in parasternal long axis; endocardial borders of the basal anterior septal and of the basal posterolateral walls needed to be clearly visualized to measure LVIDD according current guidelines and previous established experience.
Cardiac MRI was performed with a 1.0-T scanner (Magnetom Harmony Maestro Class, Siemens Healthcare, Forchheim, Germany). After orthogonal scout scans, electrocardiographic-triggered cine bidimensional steady-state free precession images were acquired during multiple periods of breath-holding in 2-, 4-, and 3-chamber long-axis views. Subsequently, using 2- and 4-chamber views as a guide, a stack of multislice images in short-axis view were obtained covering the LV from base to apex (slice thickness = 8 mm; gap = 0%; typical matrix = 350 × 350 mm or smaller). ARGUS system (Siemens Healthcare) was used to measure LV EDV (volumetric EDV) and ESV from LV short-axis views following established standard procedures. Automated detection of LV endocardial contours in short-axis images was initially used and then manually corrected when needed to obtain best fitting with blood-myocardial interface. LVIDD was measured on a short-axis slice of the LV at the tips of the mitral valve leaflets as the distance between anterior septum and the posterolateral wall as in echocardiographic acquisitions and as used previously for intermodality quantification of LV structure and function. LV EDV was also calculated from MRI LVIDD measurement using the Teichholz formula as EDV Teich = 7 × (LVIDD) 3 /(LVIDD + 2.4) and using an equation using the Z method as described by de Simone et al: EDV Z = 4.5 × LVIDD 2 .
Standardized and recommended echocardiographic methods were used for the present study. Briefly, 2D parasternal long- and short-axis views of the LV were acquired digitally (Vivid 7, GE Ultrasound, Horten, Norway), and M-Mode with optimal cursor beam orientation in each view was also recorded during patient’s apnea. LV wall thickness and internal dimensions were measured from M-Mode recordings (10% of the cases). In case of suboptimal 2D-guided M-Mode beam orientation, 2D parasternal long-axis views were used to obtain linear measurements of LV cavity and walls according to American Society of Echocardiography recommendations (90% of the cases). As in the MRI sample, EDV was calculated from the LVIDD using the Teichholz’s formula and by Z method. Doppler SV was calculated using the aortic anulus diameter as aortic flow time-velocity integral (centimeters) at the aortic anulus × aortic anulus area (square centimeters, computed as 3.14 × aortic anulus radius in centimeters) ; aortic anulus diameter was measured between the insertion points of the aortic valve leaflets (inner edge to inner edge) in parasternal or apical long-axis views, considering the largest diameter, with color flow mapping to enhance tissue-blood interfaces if needed. Transaortic flow was assessed by pulsed-wave Doppler using a 6-mm sample volume placed on aortic anulus in apical 5- and 3-chamber view; the systolic transaortic Doppler spectrum with highest peak velocity was considered. Doppler-EF was calculated as described earlier : Doppler EF = Doppler SV/EDV. The LV sphericity index was assessed as LVIDD/LV diastolic long axis in apical 4-chamber view, with greater values indicating a more spherical LV.
EF by planimetry was computed by obtaining LV planimetry at end-systole and end-diastole from apical 4- and 2-chamber views as described previously ; EDV and ESV volumes were calculated by the biplane modified Simpson’s rule, and EF was derived as 100 × (EDV−ESV)/EDV.
RT3D echocardiography was also used to assess LV EF (RT3D-EF) in the echocardiography sample. Briefly, a full-volume image for 3D analysis was obtained over 4 cardiac cycles using a matrix-array transducer (BT 8.0, General Electric) from apical views. Semiautomated LV border detection assisted the definition of LV volumes by a point-and-click method. U-shaped lines passing through the LV apex were traced from landmark points set at the medial, lateral, inferior, and anterior mitral annular corners from apical views and from the posterolateral to the anterolateral mitral annular corners in apical long-axis vies (TomTec 4D LV analysis V 2.0). The endocardial border was tracked automatically throughout the cardiac cycle, and a frame-by-frame point-and-click manual correction was performed in all cases by the operator to optimize the endocardial tracking for all the segments. In general, in our experience, the curved endocardial border after the manual correction was larger than that determined initially by the automatic procedure due to the tendency of the automatic speckle-tracking-guided procedure to overestimate the volume of the LV trabeculae and thus underestimate LV volumes. LV volumes were subsequently identified automatically as the largest and the smallest volumes and EF.
Severe segmental wall motion abnormalities were defined on the basis of the presence of at least 2 akinetic or dyskinetic near segments.
Data in tables are mean ± SD. Intraclass correlation coefficients were used as index of reproducibility. Bland-Altman plots were used to exclude systematic intermodality and/or intramodality errors when comparing a method versus a gold standard. The equation predicting EDV from LVIDD was derived using MRI data set using the curve estimation procedure, which produces curve estimation regression statistics and related plots for 11 different curve estimation regression models (linear, logarithmic, inverse, quadratic, cubic, power, compound, S-curve, logistic, growth, and exponential). A separate model was produced for each dependent variable; the best regression model was characterized by the highest R 2 and the least standard error of the estimate. A 2-tailed p <0.05 was used to reject null hypothesis, indicating statistical significance (SPSS ver. 9.0.1, SPSS, Inc., Chicago, Illinois).
Results
As shown in Table 1 , compared with the MRI sample, the separate echocardiographic testing sample was on average slightly older and showed wider LVIDD and lower EF; however, the frequency of severe wall motion abnormalities was comparable between the 2 samples.
| Variable | MRI (n = 39) | Echocardiography (n = 41) |
|---|---|---|
| Age (yrs) | 56 ± 9 | 59 ± 17 ∗ |
| Male patients | 29 (74%) | 24 (59%) |
| Ischemic cardiomyopathy | 10 (26%) | 13 (32%) |
| LVEF (%) | 56 ± 19 | 45 ± 18 ∗ |
| LVIDD (cm) | 5.6 ± 0.9 | 5.9 ± 1.2 ∗ |
| Severe segmental wall motion abnormalities | 12 (32%) | 13 (32%) |
| With ≥1 dyskinetic LV segment | 4 (10%) | 6 (15%) |
Compared with the volumetric EDV by MRI ( Table 2 ), EDV Teich was significantly greater, whereas EDV Z was smaller ( Table 2 ). The sphericity index did not enter the regression equation of EDV on LVIDD when included as a covariate (p >0.1). Bland-Altman plots showed the tendency of EDV Teich to overestimate (on average +6 ml, 80% confidence interval [CI] −18 to +14 ml) and of EDV Z to underestimate (on average −9 ml, 80% CI −47 to +7 ml) LV volumes compared with volumetric EDV ( Figure 1 ). Within the MRI sample, intrareader reproducibility of measurements of volumetric EDV was high (ρ = 0.98, 95% CI 0.97 to 0.99), as was the intraobserver reproducibility of measurements of LVIDD (5.7 ± 0.8 cm vs 5.6 ± 0.8 cm, p = NS, ρ = 0.96, 95% CI 0.94 to 0.98). Using MRI data, EDV was predicted by LVIDD ( Figure 2 ) with the highest R 2 value yielded by the following quadratic equation: EDV = (47 × LVIDD) + (1.74 × LVIDD 2 )−163, (R 2 = 0.97, p <0.0001, standard error of estimates = 2 ml, 90% CI −17 to +14 ml).
| Item: EDV (ml) | Mean Values | p Value vs Volumetric EDV | ρ (95% CI) vs Volumetric EDV |
|---|---|---|---|
| Volumetric by MRI (reference) | 154 ± 61 | — | — |
| By LVIDD using Teichholz’s formula | 160 ± 62 | <0.01 | 0.93 (0.89–0.96) |
| By LVIDD using Z method (EDV Z ) | 145 ± 48 | <0.01 | 0.91 (0.85–0.94) |
Consistently, in the RT3D echocardiography sample, EDV Teich was greater ( Table 3 ) whereas 2D biplane and EDV Z (164 ± 67 ml, p <0.01) were significantly smaller than RT3D EDV. No significant difference was seen between RT3D EDV and EDV estimated by the novel MRI-derived equation applied to echocardiographic LVIDD. Data were consistent also in subanalyses within participants with severe segmental wall motion abnormalities (n = 13) independent of the LV walls involved; in those with anteroseptal abnormalities (n = 9): RT3D EDV (244 ± 71 ml) did not differ from EDV estimated using the novel equation (243 ± 70 ml, p >0.5), whereas EDV z was significantly smaller (217 ± 60 ml, p <0.001) and EDV Teich was larger than RT3D EDV (251 ± 78 ml, p = 0.056); in participants with inferoposterolateral (n = 13), RT3D EDV (249 ± 81 ml) did not differ from EDV estimated by the novel equation (247 ± 76 ml, p >0.5), whereas EDV z was significantly smaller (219 ± 67 ml, p <0.001) and EDV Teich was larger (255 ± 87 ml, p <0.01) than RT3D EDV. Intraobserver reproducibility of measurements of RT3D EDV (ρ = 0.96, 95% CI 0.94 to 0.99) and of LVIDD were excellent (ρ = 0.92, 95% CI 0.89 to 0.98).
