Validation of volume measurements for fetal echocardiography using four-dimensional ultrasound imaging and spatiotemporal image correlation

Abstract

Objectives: To assess the accuracy and reliability of four-dimensional (4D) ultrasound imaging using spatiotemporal image correlation (STIC) employing three different techniques to measure volumes in vitro.

Methods: Customized miniature balloons attached to a pump system were used to mimic fetal cardiac chambers. After the balloon model had been immersed in a bath filled with viscous gel, 4D datasets were acquired and three methods were used for volume analysis: three dimensional (3D) slice method, Virtual Organ Computer-aided AnaLysis (VOCAL) and VOCAL combined with inversion mode. Accuracy and measurement error were measured as the difference between the volume measurements and the actual volumes. Intraobserver reliability was assessed by computing coefficients of variation (CV) and intraclass correlation (ICC).

Results: Measurement of 76 different volumes, ranging from 0.30 to 4.95 mL, resulted in a total of 912 measurements. The 3D slice method had a mean error of -3.3%, the inversion method underestimated the volumes with a mean error of -6.1%, and VOCAL had a mean error of -2.9%. The 3D slice method had the best agreement (95% limits of agreement (LOA), -11.2 to 4.7%), followed by VOCAL (95% LOA, -14.1 to 8.3%); the inversion mode demonstrated the worst agreement (95% LOA, -21.4 to 9.2%). All three methods were reliable with CV < 10% and ICC > 0.95.

Conclusions: 4D ultrasonography with STIC is a feasible and accurate method for calculating volumes of 0.30 mL upwards. In an in-vitro model the 3D slice method proved accurate, was the least time consuming, had the best reliability and had the smallest LOA. This method may prove useful when applied to in-vivo investigations.