Repeatability and reproducibility of fetal cardiac ventricular volume calculations using spatiotemporal image correlation and virtual organ computer-aided analysis

Abstract

Objective: The objective of this study was to quantify the repeatability and reproducibility of fetal cardiac ventricular volumes obtained using spatiotemporal image correlation (STIC) and Virtual Organ Computer-Aided Analysis (VOCAL; GE Healthcare, Kretztechnik, Zipf, Austria).

Methods: A technique was developed to compute ventricular volumes using the subfeature Contour Finder: Trace. Twenty-five normal pregnancies were evaluated for the following: (1) to compare the coefficient of variation (CV) of ventricular volumes obtained using 15 degrees and 30 degrees rotation; (2) to compare the CV between 3 methods of quantifying ventricular volumes: (a) Manual Trace, (b) Inversion Mode, and (c) Contour Finder: Trace; and (3) to determine repeatability by calculating agreement and reliability of ventricular volumes when each STIC was measured twice by 3 observers. Reproducibility was assessed by obtaining 2 STICs from each of 44 normal pregnancies. For each STIC, 2 ventricular volume calculations were performed, and agreement and reliability were evaluated. Additionally, measurement error was examined.

Results: (1) Agreement was better with 15 degrees rotation than 30 degrees (15 degrees: 3.6%; 95% confidence interval [CI], 3.0%-4.2%; versus 30 degrees: 7.1%; 95% CI, 5.8%-8.6%; P < .001); (2) ventricular volumes obtained with Contour Finder: Trace had better agreement than those obtained using either Inversion Mode (Contour Finder: Trace: 3.6%; 95% CI, 3.0%-4.2%; versus Inversion Mode: 6.0%; 95% CI, 4.9%-7.2%; P < .001) or Manual Trace (10.5%; 95% CI, 8.7%-12.5%; P < .001); (3) ventricular volumes were repeatable with good agreement and excellent reliability for both intraobserver and interobserver measurements; and (4) ventricular volumes were reproducible with negligible differences in agreement and good reliability. In addition, bias between STIC acquisitions was minimal (<1%; mean percent difference, -0.4%; 95% limits of agreement, -5.4%-5.9%).

Conclusions: Fetal echocardiography using STIC and VOCAL allows repeatable and reproducible calculation of ventricular volumes with the subfeature Contour Finder: Trace.

Figure 1

Selected VOCAL rotational steps utilizing Contour Finder: Trace for each ventricle in end systole and end diastole (A: Left ventricle in systole; B: Left ventricle in diastole; C: Right ventricle in systole; D: Right ventricle in diastole) at the level of the four chamber view (A1-D1) and the rendered image (A2-D2).

Figure 2

VOCAL rotational step (left) and rendered image (right) utilizing Manual Trace (top) and Inversion Mode (bottom) for the right ventricle in diastole.

Figure 3

Bland – Altman plot of the bias, expressed in percent, between STIC acquisitions (Mean: −0.4%, 95% limits of agreement −5.4 – 5.9). There was no correlation between the percent difference and the second measurement from STIC B (r_s=0.02, P=.11).

Figure 4

Square root transformed CV for the left and right ventricles. There was no significant difference in the mean square root transformed CV between the left and right ventricles (Left: 4.5%, 95% CI: 3.6 – 5.4 versus. Right: 3.7%, 95% CI: 2.9 – 4.5; t=1.2, p=.23; data untransformed after comparison)

Figure 5

Square root transformed CV for systole and diastole. There was no significant difference in the mean square root transformed CV (Systole: 4.5%, 95% CI: 3.7 – 5.6 versus. Diastole: 3.5%, 95% CI: 2.9 – 4.3; t=1.7, P = .1; data untransformed after comparison)