Our protocol provides clear, step-by-step instructions to help healthcare professionals carry out routine cranioplasty procedures, ranging from CT imaging acquisition, PMMA prostheses manufacturing to intraoperative flap implantation (Fig. 1). Given that a manufacturing facility is capable of printing a significant number of customized cranial prostheses without retooling, and that each printing can be performed without additional cost, establishment of a single central Manufacturing Additive Unit in low-income economies could provide effective assistance (cost-effectiveness analysis) in a timely fashion to an increasing number of patients in need of cranioplasty procedures.
** Figure 1 **
The constructive guidelines comprise 3 phases which are covered in the following manner: Phase #1: CT Imaging; Phase #2: 3D manufacturing; and, Phase #3: Intraoperative molding, fitting and adjustment of PMMA prostheses.
Phase 1: CT imaging
· Patient Population
We performed a review of 54 consecutive patients undergoing CT-based, low-cost, customized intraoperative PMMA cranioplasty implants over a 26-month period (Jan 2019 - Feb 2021) at Hospital da Restauração (HR), Recife, Pernambuco, and at Hospital Municipal Miguel Couto (HMMC), Rio de Janeiro, Brazil. Ethical and Institutional review board approvals (CAAE: 79457617.9.0000.5198 [HR/PE]; CAAE: 96263418.0.0000.5279 [HMMC/RJ]) and patient or caregiver consent for data release for scientific purpose and photographs were obtained prior to the project’s initiation.
· CT scanners
Multidetector computed tomography (MDCT) and volume computed tomography (VCT) have made it simple to obtain a volume data set that can be reconstructed as well as viewed with multiplanar reformation tools (MPR). Studies of the cranial and facial region, although viewed in both soft tissue and bone windows, render its use primarily for osseous and dental structures of the cranio-maxillofacial complex.
It is crucial for image quality to position the patient in the center of the scan field. Patients lying in supine position, head first into the gantry, should be assessed at rest, maintaining teeth occlusion, and must not swallow nor chew during the imaging acquisition process. The table height should be centered such that the external auditory meatus (EAM) is at the center of the gantry so that the desired cranial defect area lies within the boundaries of a selected field of view (FOV). This is achieved by moving the patient’s head in the FOV or moving the FOV over the patient’s region of interest (ROI). Use the lowest head rest available to get the patient’s head back sufficiently so that the occlusal plane is vertical (see Table 1 – Suppl. Files). It is easier to acquire high quality images for a particular ROI using a larger FOV. To reduce or avoid ocular lens exposure, the scan angle should be parallel to a line created by the supraorbital ridge and the inner table of the posterior margin of the foramen magnum - perform one or both of these maneuvers whenever possible. Finally, a helpful way to determine the orientation of a CT scan image is to use radiopaque markers. Remember: gantry tilt is available for sequence scanning, not for spiral scanning and that gantry tilt is not available for dual source scanners.
** Table 1 **
In the CT scanner, settings are based on image slice thickness, slice spacing, number of pixels, and grey scale. Grey scale, according to Hounsfield dimensionless units (HU), represents tissue density for prompt identification of different tissues in sequential images. Accordingly, bones appear in white, soft tissues in grey, and air cavities in black. For CT of the head, contiguous or overlapping axial slices should be acquired with a slice thickness of no greater than 1 mm (see Table 2 – Suppl. Files). Data sets, presenting maxilla and mandible for standard reference, should cover 10 mm above and below the patient’s region of interest (ROI), determined by the surgeon. Please, consult manufacturer specific protocols and medical physicist to assist in determining mode and features of the CT scanning available at your hospital facility to maximize image quality and acquisition and compare to the parameters suggested in Table 1 and in Table 2 (Suppl. Files) (links A,B).
** Table 2 **
In the current study, cranioplasty CT scanning was performed using helical/spiral scanners: Somatom Definition AS 64 slice, Simens (HR/PE); Brilliance Multislice 16 Channels, Phillips, RJ (HMMC/RJ). Contiguous ≤ 1-mm reconstructed slices were produced from the data volume. The data was then downloaded from the scanner workstations and files were saved in DICOM format (Digital Imaging and Communications in Medicine) (link C), which is standard format for management of medical information and related data, including CT images, for subsequent editing. This Protocol does not include Cone Beam CT, a volumetric method for integrating images, although these data can be processed using our methodology.
Transfer of CT images in a single array (DICOM) for use and processing in an Additive Manufacturing Unit. Due to its bigger size relative to the other standard sizes of the image file, the storage and transmission of DICOM files become one of the problems in a hospital information system with limited resources assisting patients who undergo cranioplasty. The volume itself, just over 500 slices, becomes over 250 MBytes of data – and for subsequent image processing tasks one can approximately multiply that by 4, so the computer should have > 1 GB RAM. Modern hard disks should provide read performance exceeding 25 Mbytes/sec, so anything beyond 10 sec is probably not optimal. Of importance, DICOM can hold both raw pixel data and JPEG-compressed pixel data, as well as many other formats. Compression aims to reduce file size, but, with decreasing data quality of the original image. Thus, original image files should be kept at the primary hospital facility until receipt confirmation by the Additive Manufacturing Unit (Lopes da Silva et al., 2017).
Transfer or sharing of larger medical DICOM image data-sets (≈ 250 MBytes) ensures that images conform to high quality standards. Conventional methods have been:
i) DICOM-based CDs and DVDs - burning CDs or DVDs for each patient comes with its burden of costs. The cost of delivering the CD must also be taken into account. Also, CDs and DVDs can be easily lost, misplaced or get damaged.
CD-RW (Compact Disc-ReWritable) often get scratches that ruin the data and are not recommended to be used.
ii) Cloud-based DICOM solutions, such as the GoogleDrive (Google®), OneDrive (Microsoft ®) and AppleCloud (Apple ®). Most providers have free versions of their services. Enable users to access DICOM images without installing special software on their devices. Cloud-based DICOM viewers are usually ‘zero footprint’. This means that the device somebody else use to view the images will not be affected by the viewer. This offers several advantages, among them: i) almost any device can work with cloud-based viewers; and, that, ii) most standalone DICOM viewers are compatible with only one particular type of operating system (OS), either Windows or Mac OS. Zero footprint viewers, however, work through the internet browser (such as Chrome, Firefox, or Safari) and, therefore, do not require a specific kind of OS. Summary of advantages: greater accessibility, lower costs, security and safety.
iii) Sending DICOM files via File Transfer Protocol (FTP) using WeTransfer.com services– send up to 2 GBytes in one go for free. Send up to 20 GBytes in one go and allow for storage of 1 TByte (terabyte) for about U$ 120/year (WeTransfer Pro) – equivalent to four Windows or Mac laptops with 256 GBytes of storage. There are many others FTP services freely available.
iv) The DICOM files can be also stored either in pendrives or in external hard disk drives, but the services above mentioned allow for simple and fast transferring of image files from a Hospital to an Additive Manufacturing Unit, which will demand a faster internet connection depending on the frequency and amount of data transferred.
Phase #2: 3D manufacturing
The DICOM file is received at the Additive Manufacturing Unit. At this point, the image data of the CT scanned area is a series of 2D images that need to be segmented, in order to separate the bone from other tissues, and converted into 3D mesh, to generate volumetric reconstruction (Fig. 2)
** Figure 2 **
3D modelling and printing hardware
Image processing is classified as a high performance computing task. This means that computers used for such applications must meet high requirements in terms of CPU, RAM, and graphical specifications in order to achieve optimal performance. Sufficient power supply and cooling must also be ensured for the server or workstation. We strongly recommend that the chosen hardware meets minimum requirement (Table 3 – Suppl. Files).
** Table 3 **
The image segmentation process can be performed manually, automatically or using both algorithms and filters (Abdullah et al. 2019). There are several segmentation methods and algorithms for medical images, among them the threshold base, where a threshold value is chosen using the radiodensity of different tissues to select a range of pixels, which highlights and separates specific structures in the image, making for easier identification of bones and other tissues (Fig. 2A-D).
The 3D model used in the computational simulation consists of the PMMA prosthesis and half of a skull. The cranial geometry was obtained from DICOM images of a cranium of a patient with bone defects. The 3D geometry of the cranium exhibiting defects was obtained by rendering images via InVesalius®.
InVesalius® is an in-house free open-source 3D medical imaging reconstruction software that generates a 3D image from a sequence of 2D DICOM images (CT or MRI) (Fig. 2A-D; Fig. 3A,B). It has been developed by the Renato Archer Information Technology Center (CTI-RA) under the leadership of Dr Jorge Silva, Campinas/São Paulo, Brazil, supporting 24 languages (Fig. 3B,D). The software is compatible with Windows, Linux and macOS. The application of this widely used and reproduceable software has been benefiting research institutions and public hospitals in LMICs to produce accurate STL skull models for teaching, medical training or clinical purposes over the last eleven years (Fig. 3A,B). The 3D models produced by InVesalius® 3.1v are comparable to those produced by MIMICS 17.0v software (Abdullah et al. 2019; Lo Giudice et al., 2020; link D).
** Figure 3 **
Using Magics®, the prosthetic geometry was generated in correspondence to the cranial defect. The .stl files of the skull and the prosthesis were processed through Geomagic Design X®, in which the component mesh was optimized (Fig. 2E,F). Of note, the three-dimensional finite element method (3D-FEM) is limited to at one-time use, to resolve the general geometry stress analysis of the skull and to define the optimized thickness of the cranial prostheses, and the concept applied to all 3D modeling. Afterwards, the skull and prostheses images were exported to CAD Rhinoceros® where the region of contact between components was adjusted, as well as where the solid model of the combined components was generated. Fig. 4 shows the prosthesis geometry (A), the skull geometry (B) and the combined assembly (C). As seen in Fig. 4A (left side) the prosthesis possesses variable thickness. The value in the region of least thickness is approximately 2.75 mm, the central region contains a value of approximately 3.95 mm and the thickest region is approximately 5 mm thick. The average thickness for the model is around 4 mm. Fig. 4A (right side) shows a circular region previously defined by geometry for applying force in the computational model to be analyzed using the finite element method.
** Figure 4 **
Afterwards, each component was saved through the extension.step, and the model was sent to the commercial software for finite elements, Hypermesh® for mesh optimization. Each component was built with a 2D mesh using triangular elements. Subsequently, a 3D mesh was generated from solids using first order tetrahedral elements (CTETRA). Table 4 (Suppl. Files) presents the quantity of nodes and elements, ‘backbones’ of finite element analysis (FEA), from the skull and prostheses. Fig. 5 presents the finite element mesh of the model components.
 In FEA, the model is divided into small pieces, called Finite Elements (FE). Those elements connect all characteristic points (called Nodes), that lie on their circumference. This ‘connection’ is a set of equations termed shaped functions. Adjacent Elements share common Nodes (the ones on the shared edge). Thus, the shape functions of all the Elements in the model are tied by common Nodes.
** Figure 5 **
** Table 4 **
Mechanical Properties of the materials.
The properties of the skull and prosthetic material were assumed to be linearly elastic, homogenous and isotropic for analysis of resistance. The mechanical material properties of the PMMA and cortical bone are listed in Table 5 (Suppl. Files) (Ridwan-Pramana et al., 2016, 2017).
** Table 5 **
Mechanical contact characteristics
In mechanical systems, the contact between bodies is a non-linear problem that presents certain solution difficulties. The formulation of a mathematical model that appropriately expresses the stress distribution and the displacement field is one of the main difficulties of mechanical contact between solid bodies.
In a finite element analysis, contact conditions are special classes of discontinuous constraints establishing that loadings are transmitted from one part of the model to another. In this Protocol, we used Hypermesh® software to establish a relationship between the ‘glued’ contact surfaces of the skull and the prosthesis, with the skull being the “master” and the prosthesis being the “slave”. Therefore, it is expected that the customized prostheses and the bone are in perfect contact.
Contour and loading conditions
The prostheses resistance was evaluated using three different values of static loading: 50 N, 600 N and 1200 N. The 50 N load was based on the work performed by Ridwan-Pramana and collaborators (2016) and is an approximation of the reaction force induced in the skull (≈ 5 kg) when resting on a flat surface without any other force acting on it. A 1200 N force was chosen to simulate an extreme impact condition in the prosthesis region. The 600 N load was chosen as an intermediate value between 50 and 1200 N. Each load was applied perpendicularly to a circular region defined in the center of the PMMA prostheses. The contour of half of the skullcap was fixed in order to prevent translation and rotation on an xyz-coordinate axis. Fig. 6 shows the application of the force perpendicular to the prostheses and the location of the skull fixation.
** Figure 6 **
Thickness of the cranial prostheses – ‘bench’ (modeling and simulation)
The objective of the simulation was to verify whether the thickness of the PMMA cranial prostheses would be sufficient to support loads imposed on it. Fig. 7 shows the results obtained for the Von Mises equivalent stress when the applied force in the central region of the PMMA prostheses was 50 N (A), 600 N (B), and 1200 N (C).
As seen in Fig. 7, the stress distribution in the three models demonstrated the same behavior. In addition to the central region where the force was applied, it is evident that there was a tendency for stress to concentrate on the inferior edge that corresponds to the location of least prosthetic thickness. The values maximum stress were 4,25 MPa, 22,38 MPa and 44,83 MPa for loads of 50 N, 600 N and 1200 N, respectively. According to these values, the prostheses would not fail due to flow stress, since the PMMA flow stress is 72 MPa (Ridwan-Pramana et al., 2016, 2017).
** Figure 7 **
Displacement analysis is important to verify the stability of the prostheses. Therefore, the total displacement observed in all of the conditions analyzed, remained acceptable and below 1 mm. The highest displacement was 0,83 mm and was observed in the 1200 N load simulation. The distribution of displacement levels in each model can be seen in Fig. 8. The finite element analysis showed that the thickness of the prostheses was sufficient to support loads imposed on the PMMA structure. The resistance was also ensured by the safety factor (SF) attributed to the design of the thickness of the prostheses. To properly design a structure, it is necessary to restrict the stress imposed on a material to a level that is safe and stable. Thus, it is essential that the stress generated in the analyzed structure, labeled as the allowable stress (σadm), is less than the rupture stress of the material (σrup). Among the reasons for this relationship is the fact that the load for which the structure was designed for may be different from the actual loads applied to it. Another fact is that the projected measurements of the structure may not be exact, due to errors in manufacturing (Shigley et al., 2004).
** Figure 8 **
A Safety Factor (SF) for the cranial prostheses was calculated considering the maximum loading applied to the model, which was 1200 N. For this, the maximum stress obtained was 44 MPa (σadm), considering a uniform prostheses thickness of 4 mm. Since the PMMA rupture stress is 72 MPa (σrup) (Table 5), the SF calculated according to the formula below it is 1.6, consistent with the literature (Shigley et al., 2004).
SF for the prostheses:
SF = σrup∕ σadm = 1.60
Therefore, considering the above-mentioned conditions for SF simulation, the cranial prostheses thickness of 4 mm was adopted in all prostheses produced by our group. It is worth mentioning, again, that this procedure of simulation is limited to a one-time event.
The molding of a personalized prostheses, in general, is done manually during surgical procedures (Maricevich et al. 2018, 2019). This can lead to variations in the thickness of the prostheses in addition to specific defects, such as bubbles, which can form during the polymer curing process. The person responsible for the production of the prostheses should also ensure the homogeneity of the mixture (liquid and solid components). Such factors could influence the final resistance of the cranial prostheses.
According to Saura (2014), one of the requirements for materials for craniofacial bone reconstruction is that these materials shall have considerable order of magnitude in terms of mechanical resistance and deformation as the original bone. Thus, one of the reasons for choosing PMMA lies in the fact that the resistance prior to rupture is approximately 72 MPa (Table 5 – Suppl. Files) while that of the cortical bone of the skull is in the range of 65 MPa to 130 MPa (Boruah et al., 2020; Ridwan- Pramana et al., 2016). On the other hand, the use of metal as titanium alloy prostheses with resistance of 900 MPa (Shigley et al., 2004) can result in force transmission between implant and skull, at increased risk of fracture of bone tissue around the prostheses.
Thickness of the cranial prostheses – ‘bedside’ (multiple episodes of TBI)
When evaluating the need to perform a large cranial repairment procedure (> 50 cm2), the surgeon needs to be aware of the choice of the ideal material for reconstruction, adopting a safety design to prevent fracturing the skull again after multiple rather than a single episode of head trauma – as ‘a lightning may definitely strike the same place twice’ (Fig. 9).
** Figure 9 **
From bench to bedside: proof-of-concept
Lack of complications following prosthetic replacement in a patient with subsequent episodes of TBI, under real life conditions, supported our computational strategy, using the finite element method, simulating the stress of static device placed onto the skull. From the obtained results, we confirmed that the 4 mm thickness PMMA prostheses is effective for cranioplasty. Therefore, the 4 mm-thickness is adequate to guarantee the mechanical protection of the skullcap considering the conditions imposed on the product – and also biomechanical properties similar to the bone tissue (see Fig. 12). Thus, in contrast to a range of materials available, comminuted fracture of the PMMA prostheses, associated with eventual subsequent episodes of head injury, spares the bone framework as predicted, efficiently preventing enlargement of the primary bone defect.
Autogenous bone grafts remain the gold standard for cranial reconstruction. While several problems remain that limit the wide utilization of such option, including regulatory requirements, high costs, comorbidity as well as method-specific limitations, customized intraoperative PMMA implants manufactured over the rapid prototyping molds proved to be effective and feasible. Thus, although advances in tissue engineering and biomaterials technology are expected alternatives on the long-run, the current cranioplasty Protocol represents a realistic approach that support a safe and cost-effective delivery of care.