Variables Influencing the Accuracy of 2‐Dimensional and Real‐time 3‐Dimensional Echocardiography for Assessment of Small Volumes, Areas, and Distances

Abstract

Objectives The aim of this study was to assess the validity, accuracy, and reproducibility of real‐time 3‐dimensional (3D) echocardiography for small distances, areas, and volumes. Methods Real‐time 3D echocardiography using matrix technology was performed in small calibrated tissue‐mimicking phantoms and compared with 2‐dimensional (2D) echocardiography. In a systematic variation of variables on data acquisition and analysis including different 3D workstations (manual disk summation versus semiautomatic border detection), the relative contributions of sources of errors were determined. The clinical relevance of the in vitro findings was assessed in 5 neonates and infants. Results Distance calculation was valid (mean relative error ± SD, −0.15% ± 1.2%). Underestimation of areas and volumes was significant for both 2D and 3D echocardiography (area: 2D, −7.0% ± 2.9%; 3D, −6.0% ± 2.8%; volume: 2D, −13.1% ± 4.5%; 3D, −6.7% ± 2.5%; P < .05). Adjustment of compression and gain on data acquisition (difference of the means: 2D, 11.6%; 3D, 17.9%), gain on postprocessing (3D, 3.4%), and the border detection algorithm on analysis (2D, 4.8%; 3D, 16.6%) had a highly significant effect on volume and area calculations (P < .001). In vivo, compression and gain on acquisition (3D, 19.1%) and the 3D workstation on analysis (3D, 22.2%) had a highly significant impact on left ventricular volumetry (P < .001). Conclusions Real‐time 3D echocardiography is a reliable method for calculation of small distances, areas, and volumes comparable with the size of the neonatal and infant heart. Variables influencing boundary identification during image acquisition and analysis have a significant impact on 2D and 3D area and volume calculations. Standardized protocols are mandatory to avoid these sources of error in both clinical practice and research.

Real-Time Three-Dimensional Echocardiography of the Left Ventricle—Pediatric Percentiles and Head-to-Head Comparison of Different Contour-Finding Algorithms: A Multicenter Study

“…Insights regarding the use of real-time 3D matrix transducers 15 and restrictions in spatiotemporal accuracy of RT3DE in pediatrics 16 were incorporated into the acquisition procedures. Measurement accuracy of the ultrasound scanners was validated using calibrated, static, tissue-mimicking phantoms 17 as well as moving phantoms. 16 In the core lab in Bonn, all data were reviewed and quantified.…”

Image‐Derived Assessment of Left Ventricular Mass in Fetal Myocardial Hypertrophy by 4‐Dimensional Echocardiography

“…Herberg et al 32 studied LV mass using in vitro models and magnetic resonance imaging, and their results indicated that real‐time 3D echocardiography was accurate for calculation of small distances and areas and was superior to 2D methods for calculation of small volumes with the size of the neonatal and infant heart (area: 2D, −7.0% ± 2.9%; 3D, −6.0% ± 2.8%; volume: 2D, −13.1% ± 4.5%; 3D, −6.7% ± 2.5%; P < .05). Real‐time 3D echocardiography showed significantly lower intraobserver and interobserver variability compared with 2D echocardiography (intraobserver coefficient of variation: 2D, 3.3%; 3D, 0.7%; interobserver coefficient of variation: 2D, 4.1%; 3D, 1.1%).…”

“…All of the images acquired from 4D echocardiography require manually modified tracing of endocardial and epicardial boundaries rather than the use of automated border detection because of the relatively low temporal resolution, and this process is time‐consuming. The gain, compression, and 3D workstation are still important factors for accurately analyzing the data, 32 so we hope that machines improve in resolution and the analysis software progresses in border‐tracking capabilities in the future. The frame rates used to acquire 4D images in this study were relatively low compared with current 2D modalities.…”

“…All of the images acquired from 4D echocardiography require manually modified tracing of endocardial and epicardial boundaries rather than the use of automated border detection because of the relatively low temporal resolution, and this process is time-consuming. The gain, compression, and 3D workstation are still important factors for accurately analyzing the data, 32 Comparison of normalized differences from reference mass between different stroke volumes at heart rates of 60 and 120 bpm. There was no statistically significant difference between 2 mL/60 bpm and 4 mL/60 bpm or 2 mL/120 bpm and 4 mL/120 bpm.…”

Multicenter Study Comparing Shunt Type in the Norwood Procedure for Single-Ventricle Lesions

“…However, RV volumes by 3DE were consistently smaller by 9%. Helberg 22 compared 3DE volumes to small, calibrated, tissue-mimicking phantoms. Although the authors concluded that 3DE was a reliable method for calculation of small distances, areas and volumes, 3D echo volume measurements were consistently smaller.…”