Three-Dimensional Echocardiographic Assessment of Left Heart Chamber Size and Function with Fully Automated Quantification Software in Patients with Atrial Fibrillation


Echocardiographic determination of left heart chamber volumetric parameters by using manual tracings during multiple beats is tedious in atrial fibrillation (AF). The aim of this study was to determine the usefulness of fully automated left chamber quantification software with single-beat three-dimensional transthoracic echocardiographic data sets in patients with AF.


Single-beat full-volume three-dimensional transthoracic echocardiographic data sets were prospectively acquired during consecutive multiple cardiac beats (≥10 beats) in 88 patients with AF. In protocol 1, left ventricular volumes, left ventricular ejection fraction, and maximal left atrial volume were validated using automated quantification against the manual tracing method in identical beats in 10 patients. In protocol 2, automated quantification–derived averaged values from multiple beats were compared with the corresponding values obtained from the indexed beat in all patients.


Excellent correlations of left chamber parameters between automated quantification and the manual method were observed ( r = 0.88–0.98) in protocol 1. The time required for the analysis with the automated quantification method (5 min) was significantly less compared with the manual method (27 min) ( P < .0001). In protocol 2, there were excellent linear correlations between the averaged left chamber parameters and the corresponding values obtained from the indexed beat ( r = 0.94–0.99), and test-retest variability of left chamber parameters was low (3.5%–4.8%).


Three-dimensional transthoracic echocardiography with fully automated quantification software is a rapid and reliable way to measure averaged values of left heart chamber parameters during multiple consecutive beats. Thus, it is a potential new approach for left chamber quantification in patients with AF in daily routine practice.


  • Excellent correlations of LV volumes, LVEF, and maximal LA volume between fully automated quantification software and the manual method were observed in beat-to-beat analysis in patients with AF.

  • The time required for analysis with the automated quantification method was significantly less compared with the manual method.

  • Excellent correlations were also noted between averaged left chamber parameters and the corresponding values obtained from the index beat.

  • Three-dimensional TTE imaging with fully automated software is a new potential approach to quantify left chamber size and function in patients with AF.

Because of the rapid increase in the aging population, atrial fibrillation (AF) is becoming a significant medical concern because of its increased morbidity and mortality. Although echocardiographic assessment of heart chamber function in patients with AF is of paramount importance, especially in those with heart failure, beat-to-beat variability in the mechanical parameters makes accurate assessment difficult when analyzing a single beat. Previous studies have proposed that cardiac function be evaluated by averaging values of parameters obtained over multiple consecutive cardiac cycles (five to 13 beats). However, this method is tedious and time-consuming and is infrequently performed in busy echocardiography laboratories. The advent of fully automated quantification software for heart chamber size and function could overcome this problem. Application of three-dimensional (3D) transthoracic echocardiographic (TTE) imaging is also limited in AF because 3D TTE full-volume data sets with high volume rates require multibeat acquisition, which in AF results in stitching artifacts, which preclude real-time beat-to-beat analysis. To circumvent these limitations, fully automated analytic software in conjunction with single-beat 3D TTE full-volume acquisition has been recently introduced. We hypothesized that this new automated software has the potential to reliably assess left ventricular (LV) and left atrial (LA) volumes in patients with AF.

Accordingly, the aims of this study were to (1) assess the accuracy of single-beat 3D TTE full-volume data sets to measure LV volumes, LV ejection fraction (LVEF), and LA volumes using a fully automated quantification software compared with a validated manual 3D tracing method; (2) determine beat-to-beat variability in mechanical parameters; and (3) demonstrate whether index-beat determination of left chamber mechanical parameters is closely correlated with the corresponding averaged values in patients with AF.


Study Subjects

From October 6, 2015 to February 9, 2016, we prospectively selected patients with AF who were referred for clinically indicated echocardiographic examinations. Among a total of 170 consecutive patients with AF, we could not perform 3D TTE data acquisition in 75 patients because of time constraints (the echocardiographic machine was being used for other examinations; n = 10), the expert sonographer or physician for 3D TTE acquisition was not available ( n = 17), or patients were unable to come to the echocardiography laboratory because of unstable condition ( n = 48). In addition, the automated quantification software generated wrong LV casts that could not be corrected with manual editing in another seven patients, who were excluded from the analysis. Thus, the final group consisted of 88 patients. All patients were Japanese. We did not use ultrasound contrast agent in any patient in this study. Some of this study’s subjects had been used in a different analysis reported in a previous publication. The ethics committee of the University of Occupational and Environmental Health Hospital approved the study protocol, and informed consent was obtained from all participants.

Three-Dimensional Echocardiography

Real-time 3D TTE full-volume data sets were acquired from the apical window with the patient in the left lateral decubitus position using an EPIQ 7G scanner (Philips Healthcare, Andover, MA) equipped with a fully sampled matrix-array transducer (X5-1). The gain and compression were adjusted to minimize dropout of the LV myocardial borders. The depth and sector angle were adjusted to include the entire left ventricle and left atrium. We used novel single-beat acquisition mode and acquired multiple consecutive cardiac beats ranging from 10 to 20 cardiac cycles according to the patient’s ability to hold the breath. After storing multiple consecutive 3D TTE full-volume data sets, we trimmed a single cardiac cycle and stored each individual beat separately from one another, starting with the first beat and ending with the last beat for later analysis of each individual beats. Three-dimensional TTE full-volume data sets were stored digitally for offline analysis.

Three-Dimensional Echocardiographic Analysis: Automated Quantification

Single-beat 3D TTE full-volume data sets were analyzed using fully automated quantification software (HeartModel; Philips Healthcare) that simultaneously detects LV and LA endocardial surfaces using an adaptive analytical algorithm that consists of knowledge-based identification of initial global shape and LV and LA chamber orientation, followed by patient-specific adaptation. After initiating the program, the software automatically determines the end-diastolic and end-systolic frames using motion analysis. Subsequently, the software builds end-diastolic and end-systolic 3D casts of the LV cavity and an end-systolic cast of the LA cavity, from which LV and LA volumes are derived directly without geometric assumptions. If the user is not satisfied with the automated LV and LA contours, the software allows global and regional editing on the end-diastolic and end-systolic apical four-, two-, and three-chamber cut planes derived from 3D TTE data sets. The LV casts can be edited by either changing the entire border globally (dilating or contracting the entire surface uniformly by the same distance) or editing it regionally. In contrast, the LA cast can be edited only regionally.

Fundamentally, in this study, we did not perform manual editing in each patient, because it requires additional analysis time and violates the principle of a fully automated analysis. The analysis was performed in each beat in all patients, and the averaged value was obtained.

For the determination of the index beat, we measured the preceding RR interval (RR1) and the pre-preceding RR interval (RR2) and calculated their ratio (RR1/RR2). The index beat was defined as the RR1/RR2 ratio closest to 1.0. If more than one beat had an RR1/RR2 ratio of 1.0, we selected the first beat having an RR1/RR2 ratio of 1.0 as the index beat. The index-beat parameters were compared with the corresponding averaged values.

Three-Dimensional Echocardiographic Analysis: Manual Method

To determine the accuracy of left heart chamber performance parameters obtained with the automated quantification software, we also measured the same parameters in each beat using the two-dimensional (2D) biplane method of disks extracted from the identical single-beat 3D TTE data sets (3DQ; Philips Healthcare). For LV volume measurements, nonforeshortened apical four- and two-chamber views were extracted from the 3D TTE full-volume data sets. LV endocardial borders were traced in both views at end-diastole and end-systole, and LV volumes were calculated using the biplane method of disks. LVEF was calculated using the standard formula. For the determination of maximal LA volume (LAVmax), we extracted anterior-posterior and medial-lateral long-axis views of the left atrium in the end-systolic frame from the identical data set. Care was taken to ensure that both cutting planes bisected the center of the short-axis image of LA area. Manual tracing of the LA wall on two long-axis views was performed, and LAVmax was measured using the biplane method of disks.

Image Quality Analysis

The image quality of 3D transthoracic echocardiography was assessed in apical four-, two-, and three-chamber cut planes extracted from the 3D TTE data sets to evaluate LV endocardial border visualization in the 18 LV segments. Image quality was defined as good when the LV endocardium was clearly visualized in >17 segments, fair when 15 or 16 segments were visible, and poor when ≤14 of segments were visible.

Beat-to-Beat Variability of Left Heart Chamber Mechanical Parameters

Beat-to-beat variability of the heart rate (HR), LV volumes, LVEF, and LAVmax was determined by the absolute difference between the maximal value and minimal value divided by the mean value during several consecutive cardiac cycles (percentage) in each patient.

Study Protocol

In protocol 1, we measured LV volumes, LVEF, and LAVmax using the fully automated quantification software for each heartbeat during 10 consecutive cardiac cycles in 10 patients. In all these patients, the number of cardiac cycles per 3D TTE data set acquisition was 10. For the determination of accuracy, we compared these values with the corresponding values measured in the same beats by the manual tracing method (3DQ). We also measured the time for the analysis between both methods. We recorded the time from the first beat measurement to the end of the 10th beat measurement.

In protocol 2, we measured the beat-to-beat variability of HR and left heart chamber size and functional parameters. We also compared automated quantification–derived averaged values from multiple beats with the corresponding values obtained from indexed beats in 88 patients.

Test-Retest Variability

Without any manual contour editing, the automated quantification software provides identical LV and LA contours, resulting in the generation of identical values of all left heart chamber parameters on the same full-volume data sets, and thus has intra- and interobserver variability of 0%. To determine the robustness of this method in AF, we acquired single-beat full-volume data sets during multiple consecutive beats at different times and performed automated quantification analysis in each beat during the first acquisition and second acquisition. Test-retest variability was determined as the absolute difference between the averaged value during the first and second acquisitions divided by the mean and intraclass coefficient.

Statistical Analysis

Continuous variables are expressed as mean ± SD or as median and interquartile range (IQR) according to data distribution. All statistical analyses were performed using commercially available software (JMP version 11.0 [SAS Institute, Cary, NC] and GraphPad Prism version 5 [GraphPad Software, San Diego, CA]). Intergroup differences were assessed using χ 2 or Fisher exact tests. Left heart chamber volume measurements from the two methods were compared using the paired t test. Comparison of percentage variability among four parameters was performed using repeated-measure of analysis of variance with the post hoc Bonferroni test. A linear regression analysis was used to assess the correlations of variation between the two methods. Bias and limits of agreement were assessed using Bland-Altman analysis. P values <.05 were considered to indicate statistical significance.


The median number of multiple consecutive beats acquired for the analysis was 11 (IQR, 10–12). The average volume rate of single-beat 3D TTE full-volume data sets was 19 ± 2/sec. Image quality analysis showed good image quality in 20 (21%), fair quality in 37 (39%), and poor quality in 38 (40%) subjects. Among 95 patients whose 3D TTE full-volume data sets were acquired, the automated quantification software did not work in seven patients. Thirty-six patients required manual editing, and no editing was required in the other 53 patients. Table 1 depicts the effect of image quality on LV contour tracking accuracy. Although image quality affected tracking accuracy, 20% of patients with good image quality still required manual editing. Figure 1 shows three representative cases, including a case with no editing required, a case with regional editing required, and a case in which the software did not work.

Table 1

Effect of image quality on LV contour determination

Number of patients No editing required Manual editing required Software not working
Good image quality 20 15 (75%) 4 (20%) 1 (5%)
Fair image quality 37 25 (67%) 11 (30%) 1 (3%)
Poor image quality 38 12 (32%) 21 (55%) 5 (13%)

Figure 1

Representative cases showing 2D cut planes derived from the 3D data sets with different degrees of left ventricular contour tracing. (A) A case with no left ventricular contour editing required. (B) A case that required regional manual contour editing. Arrows show the region that required contour editing. (C) A case in which the automated quantification software did not work. In this situation, manual editing could not correct the distorted left ventricular contour shape. LA , Left atrium; LV , left ventricle.

Protocol 1

All single-beat 3D TTE full-volume data sets could be analyzed using both the fully automated quantification method (HeartModel) and the manual method (3DQ) in all 10 patients. At a beat-to-beat level, significant linear correlations of LV end-diastolic volume (LVEDV), LV end-systolic volume (LVESV), LVEF, and LAVmax were noted between methods ( r = 0.91, r = 0.88, r = 0.91, and r = 0.98, respectively). Bland-Altman analysis revealed a small but significant bias for LVEDV and LAVmax between both methods ( Figure 2 ). The mean LVEDV, LVESV, LVEF, and LAVmax with automated quantification method were 114 ± 28 mL, 58 ± 20 mL, 49 ± 10%, and 109 ± 30 mL, respectively. The corresponding values using the manual method were 111 ± 19 mL, 57 ± 16 mL, 49 ± 9%, and 117 ± 35 mL, respectively. The automated quantification method significantly overestimated LVEDV ( P < .05) but slightly underestimated LAVmax ( P < .001) compared with the manual method. At the patient level, the corresponding correlation coefficients were 0.94, 0.93, 0.95, and 0.99, respectively. The time required for the full analysis in 10 consecutive beats was significantly less with the fully automated method (306 ± 8 sec) compared with the manual method (1,640 ± 260 sec) ( P < .001).

Figure 2

Linear correlation and Bland-Altman analysis of left chamber size and functional parameters measured using the fully automated quantification software and manual method. The red squares represent patients with good image quality, the orange circles represent those with fair image quality, and the blue triangles represent those with poor image quality. (A) EDV; (B) ESV; (C) LVEF; (D) LAVmax. EDV , End-diastolic volume; EF , ejection fraction; ESV , end-systolic volume; HM , HeartModel; LOA , limits of agreement.

Protocol 2

Although 36 patients required regional contour editing, we used the original contour values for the analysis. Clinical characteristics of the final study subjects ( n = 88) are shown in Table 2 . The median HR was 80 beats/min (IQR, 67–92 beats/min). HR variability in the individual patients ranged from 4 to 60 beats/min (median, 19 beats/min; IQR, 12–25 beats/min). Among 88 patients for the analysis, nine had regional wall motion abnormalities. With the use of fully automated software, no editing was required in five patients, and regional editing was required in four patients. The prevalence of the need for editing was not different between patients with regional wall motion abnormalities and those without.

Table 2

Clinical characteristics of the study subjects

Number 88
Men 60 (68%)
Age (y) 74 ± 12
Body surface area (m 2 ) 1.62 ± 0.19
HR (beats/min) 80 ± 19
Systolic blood pressure (mm Hg) 133 ± 22
Diastolic blood pressure (mm Hg) 81 ± 16
Primary diagnosis
Lone AF 32 (36%)
Valvular heart disease 27 (31%)
Hypertensive heart disease 13 (15%)
Ischemic heart disease 8 (9%)
Dilated cardiomyopathy 6 (7%)
Others 2 (2%)
ARB/ACE inhibitor 50 (57%)
β-blocker 52 (59%)
Calcium inhibitor 33 (38%)
Diuretics 47 (53%)
Digoxin 8 (9%)
Antiarrhythmic 5 (6%)
Anticoagulation 70 (80%)

Only gold members can continue reading. Log In or Register to continue

Apr 17, 2018 | Posted by in CARDIOLOGY | Comments Off on Three-Dimensional Echocardiographic Assessment of Left Heart Chamber Size and Function with Fully Automated Quantification Software in Patients with Atrial Fibrillation

Full access? Get Clinical Tree

Get Clinical Tree app for offline access