Data acquisition and image segmentation represent the essential processes for analyzing the anatomical structure of interest, consisting in obtaining a series of 2D images, and then employing them for the 3D model reconstruction [
4]. The 3D model generated from the segmentation is still not suitable for printing. An accurate phase that involves one or more CADs software is necessary to obtain the final printed 3D model.
A workflow for the 3D surgical guide is hereby described. It consists of sequential steps starting from image segmentation of a heart model using 2D images, ending with the actual surgical guide production via 3D printing.
2.1. Data Acquisition and Images Segmentation: 3D Heart Model Design
The data acquisition consisted in a collection of cardiovascular CT images deemed useful for reconstructing, through image segmentation, a 3D model of the anatomy of the whole heart with coronary artery tree, in order to create a 3D heart template for the building of the surgical guide. A 256-Slice CT Scanner (GE Healthcare system) was used for data acquisition of heart and intravascular structures, obtaining a good quality of images for 3D reconstruction. Moreover, the quality of the 3D models depends on the properties of the medical imaging data used. In fact, accurate models can be created when the slice thickness of the images are 0.50–1.25 mm; however, the optimal size depends on the pathology of interest [
5]. In this study, a 256-slice CT scanner was used, with a slice thickness of 0.625 mm and a size of 512 × 512 pixels, to obtain a faithful reconstruction of the heart and coronary arteries. The medical images were stored in Digital Imaging and Communications in Medicine (DICOM) format, which is the standard for storing and transmitting patient data.
Following data acquisition and clinical consideration that confirmed the presence of stenosis to be treated by CABG surgery in different coronary arteries, the second step involved the reconstruction of the 3D heart model based on image segmentation. The DICOM CT images were transferred and manipulated in 3D Slicer (Brigham and Women’s Hospital, Boston, MA, USA).
A semi-automated approach, generating a 3D volume from all areas of the images that respected the threshold range imposed, was used. To ensure quality, subsequent refining was performed manually slice-by-slice.
Three stages of image segmentation were used for reconstruction of the 3D heart model.
The first stage consisted in whole heart surface reconstruction, without consideration of the coronary artery tree structure. A threshold range between 1.00 and 375 was used to identify the whole epicardial boundary (the green volume shown in
Figure 2A). Manual editing and image filtering were then applied to isolate the heart geometries from the surrounding soft tissues, bones or other anatomical structures, and to smoothen the surface of the heart.
For the second stage of image segmentation, a semi-automated approach to reconstruct the first tract of the aorta and coronary arteries was used. The threshold range was set between 99 and 1700, because of the contrast agent used during image acquisition that brightens the shades of gray of the intravascular structures. However, further manipulation and manual editing were performed to mark the coronary tract which followed each stenosis. The optimal region for the operational cut was discussed with the medical staff. Accordingly, for the bypass placement, the regions of interest were defined as the tracts of the coronary arteries from 1 mm ± 0.5 mm after the distal stenosis to the last surgical approachable point. The latter, representing the distal end of the region of interest, was defined as the site where the vessel diameter reached 1.5 mm (
Figure 2B). Based on these regions, the operative holes were designed on the 3D surgical guide (
Figure 2B).
The third stage of image segmentation was to apply a merging logic operator to obtain the final segmentation, resulting in a combination between the segmentations above (
Figure 2C). The final model was converted into a standard tessellation language (STL) file for 3D printing. This file was generated through complex algorithms such as interpolation, which combines the segmented regions of interest into a 3D model.
The generated STL file required additional optimization and refinement before being ready for use as a template-model for the reconstruction of a 3D coronary artery tree surgical guide. Eventual imprecisions in the segmentation process, leading consequently to defects in the obtained 3D model, such as holes and mesh defects, required therefore further post-processing manipulation.
2.2. 3D Surgical Prototype Design
Meshmixer (Autodesk Inc., San Rafael, CA, USA) was used as 3D mesh and modelling CAD software, in which STL files were exported to remove segmentation imprecisions. After optimization of the 3D heart model, a virtual prototype of the surgical guide was designed.
After the refinement of the regions of interest and 3D heart model smoothing (
Figure 3), manual selection of the regions of interest (surgical holes) was performed in order to create the surgical guide shell. After an extrusion, the surgical guide with the inner surface fitting with the surface of the heart template was obtained.
To ensure a stable placement of the surgical guide on the heart, reference points were chosen. These were considered as points where the guide could be positioned on the organ to ensure its correct position. The aortic ring was considered as a reference point on the top part of the organ, with the apex as a bottom reference point. A series of structure links, with a perforated design, were added to connect the upper and the lower part to the regions of interest.
The first traced reference point was the aortic ring, defined as a surrounding structure which covers the antero-lateral aortic annulus in half circumference, capable of supporting the entire structure of the surgical guide in the upper part. The operative structure of the surgical guide was designed to cover six branches of coronary arteries: right coronary artery (RCA), including a tract of posterior descending artery (PDA), left circumflex (LCX), left anterior descending (LAD), including diagonal 1 (D1) and diagonal 2 (D2).
The perforated interlinks represented the structures that connected the aortic ring to the surgical holes and also connected the distal end of the surgical holes to the apex. The apex represented the second reference point for structure stability; it was designed to cover the heart apex with a concave shape. The surgical holes were designed upon the regions of interest. In particular, assuming a coronary artery width of 2 mm, the hole was calculated with a 2 ± 0.5 mm margin on both sides of the artery, with a total hole width of 6 ± 0.5 mm. The length of the surgical hole was chosen according to the size of the region of interest and customizable based on clinical judgement.
Once the shell of the entire model was created, a thickness was applied using the Select > Edit > Extrude function in Meshmixer. The “Offset” value was set to 2.5 mm and determined the total surgical guide thickness, while the “Direction” type had to be set to “Normal” in order to apply the extrusion uniformly.
Different sections of the surgical guide had an increased diameter to avoid breakage points, such as: (1) the aortic ring with a total thickness of 5.5 mm ± 0.5 mm; (2) the edges of the surgical holes (0.5 mm higher than the structure thickness) and (3) the apex for a total thickness of 3.5 mm ± 0.5 mm with a diameter of 30 mm. Once the surgical guide was defined, smoothing and other functions were applied to create a proper design of the 3D model. For the link region which connects the aortic ring and the apex, a perforated design was chosen, with a series of holes each 2 mm in diameter, at a distance of 10 mm from each other.
The “make solid” operation by Meshmixer, was used to create the surgical guide STL file to be printed, correcting structural errors as shown in
Figure 4.
2.3. 3D Printing
The 3D heart model was printed to real scale as a phantom to test the surgical guide fitting. The printing process selected to realize the phantom was the fused deposition modeling (FDM), using a Prusa imk3 3D printer, due to its low cost and ease of use [
5]. Polylactide (PLA), a rigid material, was used. The phantom, the heart reproduction with a patient-specific region of interest, was only used to test the fit of the surgical guide.
The Stratasys Object260 Connex1 3D printer was then used to realize the surgical guide. The model was printed in single material mode and the printer was set for glossy printing, in order to obtain a more resistant layer [
6]. PolyJet was employed for the printing process, representing one of the most recent rapid prototyping processes available on the market. It is a hybrid between selective hardening and drop deposition. The most important component is the printing head: it applies a liquid compound of reactive monomers and oligomers which polymerize in response to ultraviolet light. The application of the material to the tray is granted by the piezo-electric printing head which injects the liquid compound onto the metallic tray. After the injection of each layer, the tray moves downward one layer in thickness, and then the process is repeated with the next layer [
7]. The thickness of each layer measured 16 nanometers; therefore, this process can be ranked among the most accurate processes and the device can be considered as one of the fastest devices for rapid prototyping. Thus, the small layer thickness ensured the manufacture of a model with very smooth surface and small details. The material chosen for this research study was a flexible, transparent and biocompatible material from Stratasys: MED625FLX™ (Stratasys Inc., Eden Prairie, MN, USA) [
8].
Material, support material and printing mode were defined before starting the printing. When all settings were selected, the printing time was calculated by the software, accounting for the object position on the tray.
After completion of the printing process, the surgical guide was encased in the SUP706 (FullCure705) (
Figure 5A), a gel-like support material which had to be carefully removed to expose the delicate printed portions. To remove support material a high-pressure water jet was used: the Objet WaterJet cleaning unit, also marketed by Stratasys (
Figure 5B). The cleaning time varies for each printed model, according to its design.