Fusion of 3D ultrasound images of the fetal femur improves boundary definition and volume measurement

Objective: To combine multiple 3D volumes of the same fetal femur into one composite image data set using image registration and wavelet-based fusion. Fused and single data sets were compared in terms of image quality and femur volume (FV) measurement repeatability. Method: In healthy pregnant volunteers, six volumes of the same femur were acquired and fused into a composite data set. Image quality scores were given to the fused and single data sets by an independent assessor in a blinded fashion; repeatability of FV measurement was assessed using coefficients of variation (CV), intraclass correlation coefficients (ICC) and Bland-Altman plots. Results: Fusion was successful in 24 out of 25 cases. Median image quality score was 7/10 in fused data sets, compared to 6/10 in single data sets (p = 0.096). Repeatability of FV measurement was better in fused data sets (intraobserver CV 4.6% and ICC 0.987; interobserver CV 4.9%, ICC 0.985) compared to single ones (intraobserver CV 5.8%, ICC 0.977; interobserver CV 10.0%, ICC 0.931). The measured FV was significantly higher in fused data sets (mean FV 1.7 vs. 1.3 ml, p < 0.001). Conclusion: Image registration and wavelet-based fusion can improve image quality and FV repeatability; it also results in an increased FV measurement.