Segmentation. The blood pool is highlighted using “thresholding” based on gray values in orthogonal views (a–c). An “offset” layer is created on the surface of this model (d) which is then hollowed, leaving a representation of the intracardiac anatomy
Optimization of a 3D Virtual Model
Once an initial version of a 3D model has been created, multiple steps may be undertaken for virtual model optimization. These steps will ensure a good quality 3D printable model. The 3D virtual model represents a polygonal mesh structure that consists of a collection of vertices, edges, and faces defining the shape of the object. In a triangular mesh, the polygons are triangles of different shapes and sizes organized to create the surfaces of the cardiac model (Fig. 9.2). A high-quality mesh consists of the least of number of triangles needed to define the topology of the 3D structure. In order to simplify the mesh, the number of triangles can be reduced by merging, and the sizes and shapes of the triangles are made more consistent throughout the surface. The final optimization of the 3D model consists of multiple steps, such as cropping, smoothing, and removal of free-floating parts.
Polygonal mesh. The structure of the 3D virtual model is comprised of a triangular mesh
After a gross 3D virtual file has been created, it must be cropped so that the anatomy of interest is best visualized. In order to create a more “true-to-life” model, the end of the vessels can be cropped to represent the patent vessel lumen. This can be done for the systemic and pulmonary veins, systemic and pulmonary arteries, and any other accessory vessels in a whole-heart model, and this approach results in a more realistic appearing 3D cardiac model. In a patient with a DORV, the most significant relationships that can be demonstrated on a 3D model for presurgical planning are those between the VSD, great arteries, and ventricles (Fig. 9.3). The plane with which these relationships can be simultaneously demonstrated will vary from patient to patient. It is not necessarily true that this can be done using a single plane which simply cuts through the heart along one angle, but instead there may be a combination of planes at different angles used to reveal the anatomy from various aspects. The goal of visualizing these spatial relationships is to assess whether the patient is a candidate for a two-ventricle repair or for a single-ventricle palliation. Although this determination rests on multiple factors, one of the most important considerations is whether oxygenated blood flow can be rerouted to the aorta in an unobstructed manner. In patients with a subaortic VSD, the distance between the tricuspid valve and the pulmonary valve predicts the feasibility of a two-ventricle repair, as this is the region through which a LV to aortic baffle pathway will be created. [7–9]. As 3D models become more commonly used in patients with a DORV, it may be possible that 3D quantification using 3D printed cardiac models will improve and help to predict the optimal surgical approach for a given patient.
Cropping. The virtual model is cropped to best demonstrate the internal anatomy of interest. a The right and left ventricular free walls have been removed (dotted lines). b The model is rotated to allow viewing from the RV aspect with the VSD en face. The pathway from the left ventricle through the VSD to the aorta is of importance for presurgical planning (red arrow)
Smoothing of the virtual 3D model allows for an improvement in the overall appearance of the model. The surface of the virtual model is made less coarse without changes in the representation of the 3D anatomy. Laplacian smoothing is one algorithm by which a triangular mesh can be smoothed. It entails reassigning the position of each vertex in a mesh based on the surrounding points (Fig. 9.4). This step can often be completed with commercially available 3D post-processing imaging software or with an open-source software such as Meshlab® (Pisa, Italy). In order to complete this task in Meshlab®, the 3D file should first be imported, using the “Import Mesh” function under the File menu. Once the file has been imported, the Laplacian smooth function should be utilized under the “Smoothing, Fairing, and Deformation” option from the Filters menu. The smoothing steps auto-fills to 3 but can be changed based on preference. It is advisable to start with one smoothing step as one of the drawbacks of using the Laplacian smoothing function is that the model shrinks with each additional iteration.