Fetal biometry: how well can offline measurements from three-dimensional volumes substitute real-time two-dimensional measurements?
Sarris I., Ohuma E., Ioannou C., Sande J., Altman DG., Papageorghiou AT.
OBJECTIVES: To assess the feasibility, accuracy and reproducibility of manipulating three-dimensional (3D) volume sets in order to reconstruct optimal two-dimensional (2D) planes for fetal biometry throughout gestation and compare them with those derived from real-time 2D scanning. METHODS: Sixty-five fetuses were evaluated at a gestational age of 14-41 weeks. For each fetus a duplicate set of seven standard fetal measurements was taken by an experienced operator using 2D ultrasound and then 20 intentionally suboptimal 3D volumes from different predefined angles were captured and stored. These were manipulated and measured. The time taken to complete a full scan, with both 2D and 3D ultrasound, was recorded. All measurement differences were expressed as gestational age-specific Z-scores. For all comparisons Bland-Altman plots were used and limits of agreement were calculated. The means and variances of the measurements were tested with a paired t-test and Pitman's test for differences in variance, respectively. The difference between the time taken to perform a 2D and a 3D scan was tested using the Wilcoxon signed-ranks test. RESULTS: Mean agreement between 2D and 3D ultrasound measurements was good, with no statistically significant differences (i.e. no systematic error) unless the head was facing anteroposteriorly, or the long axis of the femur was at 60-90° to the transducer. The variance (random error) for 3D measurements was similar to that for 2D measurements. Planes from some volumes could not be extracted (7% for head circumference, 9% for abdominal circumference and 11% for femur length). The median time required to perform a full fetal biometric scan was significantly higher for 3D than for 2D (3:04 min vs 1:57 min, respectively; P < 0.001). CONCLUSIONS: Fetal measurements derived from 3D volume acquisitions exhibited good agreement with those obtained by real-time 2D scanning, with no extra systematic or random error. However, they were slower to obtain, not all volumes were amenable to extraction of planes and measurements that came from a head facing anteroposteriorly or that were obtained with the long axis of the femur at 60-90° to the transducer were systematically smaller.