Purpose or Case Report: Developmental dysplasia of the hip (DDH) is a common congenital problem affecting up to 3% of the the population. If untreated, DDH may lead to hip dislocation and premature osteoarthritis. Current DDH diagnosis is highly operator-dependent as it relies on 2D ultrasound. 3D ultrasound offers more complete, and potentially more reliable, imaging of infant hip geometry. However, it is unclear whether 3D ultrasound images, noisy with artifacts and reconstructed by proprietary algorithms from non-parallel beams, give accurate 3D shape information. We sought to validate the fidelity of acetabular surface models obtained by 3D ultrasound by comparison with those obtained concurrently from MRI.
Methods & Materials: 3D ultrasound and MRI scans were performed on the same day in 20 infants aged 0-6 months old. Coronal 3D MEDIC images (1 mm slice thickness) were obtained in a 1.5T Siemens scanner. 3D ultrasound was performed using a Philips VL13-5 probe. Acetabular surface models were generated for 40 hips from both MRI and 3D ultrasound data sets using semi-automated tracing software. Three observers traced the MRI and 3D ultrasound bony surfaces to assess the inter-observer variability in surface model generation. To evaluate the inter-modality variation between the models, the root mean square (RMS) distances between 3D ultrasound and MRI models were calculated using Amira software. This involved an initial manual transformation to align 3D ultrasound and MRI surfaces, followed by an automated iterative closest point rigid registration to minimize RMS distances between the two surfaces. Registrations were repeated on a subset of surface models to determine the reproducibility of the RMS distance calculations in Amira.
Results: Inter-modality variability was minimal on inspection and was confirmed with a mean RMS distance of 0.4+/-0.3mm. A 95% confidence interval for the mean RMS distances was <1mm for all comparison sets. Intra- and inter-observer mean RMS distances were significantly smaller for 3DUS than for MRI (P<0.05).
Conclusions: Acetabular geometry was reproduced by 3D ultrasound surface models within 1 mm of the corresponding 3D MRI surface model, and the 3DUS models were more reliable. This validates the fidelity of 3D ultrasound modelling and encourages future use of 3D ultrasound in assessing infant acetabulum anatomy, which may be useful to detect and monitor treatment of hip dysplasia.