3D printing of fetal heart using 3D ultrasound imaging data

Abstract

Conventional two-dimensional (2D) ultrasonography is the gold standard for diagnosing congenital heart disease (CHD). However, characterization of complex CHD may be limited by 2D imaging methods that lack crucial spatial information, as these images are generated as individual ‘slices’ without orientation and with a limited field of view lacking depth, and therefore they are not ideal for three-dimensional (3D) modeling and printing. While approximately 90% of 3D printed models of the cardiovascular system are derived from computed tomography (CT) or magnetic resonance imaging (MRI)1, these modalities are used sparingly in clinical practice. Advances in 3D and four-dimensional (4D) ultrasound have expanded the capabilities for assessing the fetal heart and have paved the way for creating ultrasound-derived 3D printed models. Medical 3D printing applications of 3D ultrasound-derived imaging data have been demonstrated for cardiac models of valvular and septal CHD2. Olivieri et al. verified the accuracy of models created from 3D ultrasound data of CHD patients3. Currently, while 3D ultrasound-derived 3D prints have been used to produce physical models of whole fetuses4 and for surgical rehearsal prior to fetoscopic repair of spina bifida5, the majority of prints have been created for fetal structures with a rest period, such as the face or the back. Here, we demonstrate the feasibility of 3D printing of a small fast-moving fetal anatomical structure without a rest period, i.e. the fetal heart, using only 3D ultrasound data. To create a 3D model of a fetal heart, 3D fetal ultrasound data were obtained from a Voluson E10 ultrasound machine (GE Healthcare, Chicago, IL, USA). The dataset was obtained from a set of sample image scans provided by GE. A 3D/4D GE RM6C ultrasound transducer was used to scan the heart of a fetus at 27 weeks' gestation, using spatiotemporal image correlation (STIC) settings. The dataset was exported directly from the Voluson E10 system as a Cartesian.vol file, and imported into post-processing software Mimics (Research Edition 19.0, Materialise NV, Leuven, Belgium) for segmentation. During 3D segmentation, the boundaries of the image data were verified by a sonographer (Figure 1a). A threshold value was chosen to include the blood pool, and the unwanted voxels from the mask were removed by cropping. Interpolation, to apply the threshold values to multiple slices at a time, and refinement were then carried out to complete the segmentation. 3D objects were then generated, smoothed and subsequently exported as STL files. The exported STL files were optimized using Meshmixer (Autodesk Inc., San Rafael, CA, USA) (Figure 1b,c). Finally, the models were printed on a 3D printer (Connex3 Objet260, Statasys, Eden Prairie, MN, USA), using Tango Plus material and SUP706 support material (Figure 1d). In summary, 3D ultrasound is a safe, cost-effective and accessible imaging modality that does not require the use of harmful radiation, contrast media or anesthesia, thus being safer to use than CT and MRI in vulnerable fetal, neonatal and pediatric patients. We created proof-of-concept 3D printed models derived from 3D fetal cardiac ultrasound data. Customized 3D printed models of the fetal or pediatric heart may complement 2D data to improve management and procedure planning, as well as patient outcomes, in CHD. J.L.M. is supported by the Fetal Health Foundation 2015 Brianna Marie Grant. S.A.C is supported by the Vesalius Trust for Visual Communication in the Health Sciences Alan W. Cole Scholarship. C.S.O. is supported by the American Heart Association Predoctoral Fellowship (2017) and Tan Kah Kee Foundation Postgraduate Scholarship (2017). We thank Maureen Schickel (Materialise) for technical assistance, Sarah Millard, RDMS (Center for Fetal Therapy) for verifying the boundaries of the image data, CiCi McShane (Center for Fetal Therapy) for data export and GE Healthcare for providing the fetal heart 3D ultrasound dataset.