This was a prospective study. Forty-two foetuses diagnosed with VRs and subject to cardiac ultrasound from May 2018 to December 2020 were enrolled in this study, with gestational weeks ranging from 21 to 31 weeks. Exclusion criteria comprised cases with poor 2D image quality, too much or little amniotic fluid, or foetal arrhythmia. As an initial step to create 3D models, conventional cardiac ultrasound and Spatio-temporal image correlation(STIC) volume data of the foetus, including 2D images, and qualified volume data were saved for subsequent analysis and postprocessing. It is worth noting that during STIC volume data acquisition, the 3-vessel trachea or aortic arch coronal view should be selected as the initial view to show as many brachiocephalic arteries as possible and avoid shadows and motion artefacts.
The volume data were exported as Cartesian.vol files and loaded into Mimics (version 19.0, Materialise, Leuven Belgium) for postprocessing. Threshold segmentation of the upper mediastinal region of interest was performed, during which the image boundaries were confirmed by an experienced sonographer. The minimum threshold was set to 0, and the maximum value was set to 65–110. Then, manual interactive segmentation was applied to remove noise signals and adjacent structures layer by layer. Objects were generated and exported to 3-MATIC (Version 11, Materialise, Leuven Belgium) using the following steps: repairing, wrapping, smoothing, and adding a layer of surface with 0.5 mm thickness based on the vascular model. The 3D digital models of the vessel and its branches, including the ascending aorta, aortic arch, aortic arch branches, ductus arteriosus, descending aorta, and pulmonary artery, were generated and subsequently saved as standard tessellation language(STL)format files.
The chosen STL files were printed with a Form2 3D printer using stereolithography printing technology in photosensitive resin material, which is a white opaque, hard and inelastic material. All models were 1:1 in size and scaled with a 3-fold factor to provide a deeper view of such small-sized cardiovascular structures. For a clearer display of the relationship between blood vessels and the trachea, a 3D-printed model of the trachea and bronchus was also constructed for this study. The workflow of 3D printing is shown in Fig. 1.
Quantitative Assessment: Analysis Of 3d Model Accuracy
Vernier callipers were used to measure the 1:1 scale printed models, and the measurements were compared with conventional two-dimensional echocardiographic measurements and analysed for consistency. Three measurement sites were selected for each case according to the characteristics of the case and feasibility of measurement. The measurements obtained included the diameter of the aorta (AO), pulmonary artery (PA), ductus arteriosus (DA), left pulmonary artery (LPA) and right pulmonary artery (RPA). The accuracy of the 3D modelling was evaluated by comparing the difference in measurement between the model and image data.
Qualitative Assessment: Applicability Of 3d Models
Medical education
Forty-eight ultrasound students from our hospital's obstetrics and gynaecology ultrasound department who received standardized training as resident physicians were recruited. Briefly, 48 third-year medical students were randomly assigned to either the 2D image group (n = 24) taught with the aid of a hand-drawn schematic diagram or the 3D printing group (n = 24) taught with the aid of 3D printed models. Both groups attended the same lecture on VRs in two separate 60-minute sessions each, presenting standard slides on embryology, anatomy, and classification as well as prenatal diagnosis and postnatal management of VRs. After an initial description of each category of VRs, both groups of students were allowed to freely analyse and manipulate teaching aids (2D hand-drawn schematics or 3D printed models) throughout the lecture. Students were also free to ask questions about the models or schematics during the lecture.
To evaluate knowledge acquisition, each student answered the same multiple-choice test twice, pre- and postlecture, with a maximal score of 100 points corresponding to 10 questions. The test assessed knowledge acquisition on embryology, anatomy, etc. At the end of the course, both groups were required to complete a brief 5-question satisfaction survey to establish the perceived effectiveness of the educational method to the student.
Parent questionnaire survey
After obtaining consent, a cardiologist explain the child’s heart condition to 40 parents using a 2D hand-drawn schematic diagram. Then, the parents received the same explanation using a 3D-printed heart model. After each consultation, parents received a short survey of 4 questions rated on a 5-point Likert scale to assess their perception of the use of 3D models in prenatal consultation. We deliberately chose to keep the questions concise and direct, as the scope of this study was limited to initial impressions of understanding based on each modality.
Statistical analysis
Statistical analyses were performed using SPSS Version 22.0 (SPSS Inc., Chicago, IL). Continuous variables were described as the means ± standard deviations. To assess the agreement between ultrasound images and 3D models, a Bland‒Altman analysis was used. The data collected from the 2D image and 3D printing groups were then compared to each other using Fisher's exact test and the Mann‒Whitney U test. A chi-square test was performed for the students’ and parents’ satisfaction data. All tests were assessed at the P value < 0.05 level of significance.