Three-Dimensional Modelling of the Fetal Vasculature from Prenatal MRI using Motion-Corrected Slice-to-Volume Registration
David F A Lloyd1,2, Bernhard Kainz3, Joshua F P van Amerom1, Kuberan Pushparajah1,2, John M Simpson2, Vita Zidere2, Owen Miller2, Gurleen Sharland2, Tong Zhang1, Maelene Lohezic1, Joanne Allsop1, Matthew Fox1, Christina Malamateniou1, Mary Rutherford1, Jo Hajnal1, and Reza Razavi1,2

1Division of Imaging Sciences and Biomedical Engineering, King's College London, London, United Kingdom, 2Evelina Children's Hospital, London, United Kingdom, 3Department of Computing (BioMedIA), Imperial College London, London, United Kingdom

Synopsis

The diagnosis of potentially life-threatening vascular abnormalities in the fetus can be difficult with ultrasound alone. MRI is one of the few safe alternative imaging modalities in pregnancy; however to date it has been limited by unpredictable fetal and maternal motion during acquisition. We present six antenatal cases, four with important structural congenital heart disease, in which we employed a novel algorithm for motion-corrected slice-volume registration, producing a navigable 3D volume of the fetal thoracic vasculature. The anatomical findings in each case were then correlated to fetal echocardiographic findings, and finally displayed as interactive surface rendered models.

Purpose

The antenatal diagnosis of important vascular abnormalities allows for both better informed parental counselling and the planned provision of potentially life-saving care after birth. The fetal vasculature can however be difficult to visualise with ultrasound,1 and the prenatal diagnosis of conditions such as coarctation of the aorta may be based solely on subtle anatomical markers which may be difficult to detect before birth.2 MRI is one of the few alternative imaging techniques which is safe to use in pregnancy,3 but it has been limited technically by unpredictable gross fetal and maternal movements.4 We sought to apply a novel motion-corrected slice-to-volume registration algorithm to antenatally acquired MRI data from fetuses with and without congenital heart disease, in an attempt to improve in utero visualisation of the thoracic vasculature.

Methods

Multi-slice MRI sequences were planned to obtain full coverage of the fetal thorax in multiple orthogonal orientations, using overlapping 2D single-shot fast spin echo sequences (ssFSE) (Philips, 1.5T, TR = 15000ms, TE = 80-100ms, flip angle = 90 degrees, field of view = 350 x 350mm, voxel size = 1.4 x 1.4 mm, slice thickness = 2.5mm). The resulting motion-corrupted stacks of ssFSE slices (figure 1A) were processed using a novel parallel super-resolution algorithm for slice-to-volume registration.5,6 This uses an iterative loop to optimise 2D/3D registration based on image intensities, incorporating edge-preserving anisotropic diffusion filtering and automatic exclusion of outlier data. To confirm anatomical accuracy, the resulting 3D volume was compared to fetal echocardiographic data by a clinician with experience in fetal cardiology. A surface rendering of the fetal vessels was then produced from the inverted dataset using Osirix™, an open-source DICOM reader. Final cropping and shading was performed using MeshLab (Visual Computing Lab - ISTI – CNR, meshlab.sourceforge.net).

Results

MRI data was acquired in two normal fetuses (at 38 and 39 weeks) and four fetuses with congenital heart disease (coarctation of the aorta at 36 weeks, hypoplastic left heart syndrome at 31 weeks, tetralogy of Fallot at 23 weeks and rightward cardiac axis with bilateral SVCs at 35 weeks). Motion-corrected super-resolution 3D volumes were produced for each patient with an isotropic voxel size between 0.40 and 0.55mm, which could be navigated in three dimensions using standard multi-planar reconstruction software (Osirix™, figure 1B). In each case, the reconstructed images accurately represented the general vascular anatomy in terms of connections and spatial relationships within the thorax when compared to fetal echocardiography, as demonstrated in figure 2. Surface renderings of the vascular data are shown in figures 3 and 4.

Discussion

Many forms of congenital heart disease are associated with important vascular abnormalities, which can be life-threatening in the immediate postnatal period if undetected. By combining prenatal MRI with advanced motion correction algorithms, we have demonstrated a novel and robust method of obtaining detailed imaging of the fetal vasculature whilst compensating for fetal and maternal motion - previously a major limiting factor to more widespread adoption of prenatal MRI.4 In conjunction with conventional modalities, these advanced techniques offer the potential both to improve the antenatal detection of important congenital vascular lesions, and to enhance our understanding of this complex group of abnormalities.

Conclusion

We have demonstrated a novel technique for imaging the fetal vasculature, which offers a level of detail that is yet to be described in the literature. Further work is ongoing to validate these techniques and establish their prognostic value when combined with established antenatal imaging modalities.

Acknowledgements

This work was supported by the iFind Project (Wellcome Trust IEH Award 102431). In addition, the authors acknowledge financial support from the Department of Health via the National Institute for Health Research (NIHR) comprehensive Biomedical Research Centre award to Guy's & St Thomas' NHS Foundation Trust in partnership with King's College London and King’s College Hospital NHS Foundation Trust.

References

1. Simpson JM. Impact of fetal echocardiography. Ann Pediatr Cardiol. 2009;2(1):41-50.

2. Jowett V, Aparicio P, Santhakumaran S, Seale A, Jicinska H, Gardiner HM. Sonographic predictors of surgery in fetal coarctation of the aorta. Ultrasound in Obstetrics & Gynecology. 2012;40(1):47-54.

3. Reddy UM, Abuhamad AZ, Levine D, Saade GR. Fetal Imaging Executive Summary of a Joint Eunice Kennedy Shriver National Institute of Child Health and Human Development, Society for Maternal-Fetal Medicine, American Institute of Ultrasound in Medicine, American College of Obstetricians and Gynecologists, American College of Radiology, Society for Pediatric Radiology, and Society of Radiologists in Ultrasound Fetal Imaging Workshop. J Ultrasound Med. 2014;33(5):745-757.

4. Malamateniou C, Malik SJ, Counsell SJ, et al. Motion-compensation techniques in neonatal and fetal MR imaging. AJNR Am J Neuroradiol. 2013;34(6):1124-1136.

5. Kainz B, Malamateniou C, Murgasova M, et al. Motion corrected 3D reconstruction of the fetal thorax from prenatal MRI. Med Image Comput Comput Assist Interv. 2014;17(Pt 2):284-291.

6. Kainz B, Steinberger M, Wein W, et al. Fast Volume Reconstruction From Motion Corrupted Stacks of 2D Slices. Medical Imaging, IEEE Transactions on. 2015;34(9):1901-1913.

Figures

Figure 1: Multi-planar reconstruction of a multi-slice fast spin echo (ssFSE) MRI stack in a normal fetus (A: upper three panels). The reconstructed planes orthogonal to the acquisition plane show the extent of fetal movement between slices. Following the application of a novel motion-correction algorithm to multiple ssFSE stacks, a fully navigable 3D volume of the fetal thorax has been produced (B: lower three panels), providing detailed views of the extracardiac vasculature.

Figure 2: Standard fetal “three vessel view” on ultrasound (A) and reconstructed MRI (B). P = pulmonary artery, Ao = aorta, S = superior caval vein, t = trachea.

Figure 3: Surface rendering of the major fetal vessels from motion-corrected data in a normal fetus at 39 weeks gestation. SCV = superior caval vein, BCA = brachiocephalic artery, LCA = left common carotid artery, LSA = left subclavian artery, IV = innominate vein, Ao = aorta, AD = arterial duct, DAo = descending aorta, PA = pulmonary artery, LPA = left pulmonary artery.

Figure 4: Surface rendering of the major fetal vessels from motion-corrected data in a fetus with hypoplastic left heart syndrome at 31 weeks gestation. Note the diminutive ascending aorta (*). SCV = superior caval vein, BCA = brachiocephalic artery, LCA = left common carotid artery, LSA = left subclavian artery, IV = innominate vein, Ao = aorta, AD = arterial duct, PA = pulmonary artery, DAo = descending aorta, LPA = left pulmonary artery, RUPV = right upper pulmonary vein.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0413