Virtual Surgical Planning and Pre-operative Simulation for Maxillofacial Reconstruction with 3D Printing Technology in Tumor-affected Facial Deformity

Background: The high anatomical complexity of maxillofacial defects caused by tumor can pose a formidable challenge for clinicians when designing an appropriate plan for surgical reconstruction. The intention of this work was to restore the complex anatomy with maximum possible facial functionality and aesthetics of the patient. Based on the medical images generated by computed tomography (CT) scan an optimal therapeutic planning for complex maxillofacial reconstruction was designed. Method: Firstly, the volumetric data sets were carefully evaluated and deeply inspected for accurate diagnosis. Regarding 3D visualization of the CT scan images 3D virtual models for regions of interest were created using a special software of 3D Slicer. Using the resulting 3D virtual models a well-defined virtual surgical planning was generated for multiple surgical procedures, including the osteotomies for bone defects, harvesting autogenous bone graft and creating a customized implant. Results: The relevant patient-specific anatomical models for real surgery were translated into the 3D printed physical models, with which the surgeons can rehearse the surgery before coming into the operating room. Precisely defined multiple surgeries for complex maxillofacial reconstruction were proposed in this research that could be transferred to the real-time surgery. Conclusions: The proposed surgical approach will be beneficial both for the surgeons and patient, including improvement in surgical precision and outcomes, reduction in operating time, as well as understanding surgical procedures and decision making etc. 3D: Three dimensional; ABS: Acrylonitrile butadiene styrene; PLA: Polylactic acid; CAD: design; CAM: Computer-aided CT: Computed tomography; DICOM: Digital and communications FDM: Fused deposition modeling; IV: Intravenous; MRI: Magnetic resonance Imaging; STL: Stereolithography or standard tessellation ROI: Region of interest; ZMC: Zygomaticomaxillary complex; DVR: VSP: Virtual planning.


Introduction
Extensive maxillofacial defects, which can be caused ether by secondary to trauma, ablative tumor resection, and refractory infection or by congenital/developmental deformities, lead to the abnormal contours of the face and functional disorders [1,2]. The face is vital for human identity, and plays a crucial role in the individual's social interaction. Thus, individuals with even minor facial deformities are at risk for experiencing social and psychological stress and thus for affecting their quality of life adversely.
Reconstruction of maxillofacial defects, especially the restoration of aesthetics and function, is always a difficult task, because the bone lesions that lead to deformities of the normal maxillofacial anatomy are complex in general [3,4]. There are several factors which are needed to be considered in the planning of these reconstructions, including the anatomic diversity of the bone defects and the complexity of the oral function based on mandibular movement, and normal facial appearance. All these aspects are important for maintaining quality of life.
The demand for performing accurate and low risk surgery, which is still a great challenge for the surgeons at the present time, has led to the discovery of computer-based surgical planning [5,6]. Nowadays computer-assisted surgery is becoming more and more important in the field of oral and maxillofacial surgery, because of the possibility of simulating almost every scenario imaginable with a high accuracy [7].
Today, the medical imaging allows us to see the structures inside the human body in great detail noninvasively that plays a crucial role throughout the entire clinical applications from medical and laboratory research to diagnostics and therapeutic planning.
In current, medical imaging is increasingly utilized in clinical practice and research settings to qualitatively and quantitatively characterize normal anatomy and physiology to make it possible to identify abnormalities [8]. In recent years, 3-dimensional (3D) images, especially 3D reconstructions from computed tomography (CT) or magnetic resonance imaging (MRI) scan, have played an important role in preoperative evaluation [9][10][11]. It is required for accurate clinical decisions and efficient surgical procedures to achieve the rapid appreciation of the anatomical and pathological information contained within medical images from the patients.
With increasing availability of high-resolution imaging modalities, technologies for 3D visualization of medical imaging data sets are rapidly gaining importance in medicine.
Currently, the great challenges are not only how best to display the volumetric data sets generated by CT and MRI scan, but also how accurate to represent those data into the tangible, physical 3D objects [12]. 3D printing is a method of fabrication that allows for the creation of graspable 3D objects from 3D imaging data.
Recently, 3D printing of anatomic models based on CT and MRI scans of patients has been successfully used for detailed preoperative surgical planning and simulation before its performance in clinics [13]. Advanced digital techniques -computer-aided design (CAD), computer-aided manufacturing (CAM) and 3D printing -have been widely used in modern medicine to improve the accuracy and quality of diagnosis and surgery [14][15][16].
The purpose of this study was threefold. The first was to evaluate the maxillofacial medical images generated by CT scan for a patient, who showed a severe facial deformity caused by tumor, for creating patient-specific, highly accurate 3D anatomical models in the context of complex reconstruction of maxillofacial bone defects. The second purpose was to develop a virtual surgical plan (VSP) for foreseeing and managing almost every imaginable scenario that could arise during the surgical procedure as well as to design the customized intraoperative surgical tools with greater accuracy. That helps doctors in taking right decisions even before performing the surgery. The last purpose was to simulate the surgery assisted by 3D printed patient-specific anatomical models as templates with customized implant and autogenous bone graft for training residents as well as experienced surgeons prior to entering the operating room, by which it will be produced more predictable outcomes and reduced duration of surgery and surgical risks significantly.
We hypothesize that the combination of more accurate VSP with 3D virtual models and preoperative simulation with 3D physical models will perfectualize an entire surgical plan for maxillofacial reconstruction, which results in improved outcome meaningfully.

Methods: Patient
A 24 year-old-male presented with a gradually enlarging, refractory swelling on the left side of his face, leading to obvious facial asymmetry. The patient denied any pain or numbness associated with this swelling on palpation. The teeth in the area of interest were non-mobile, non-painful. The histopathological analysis from incisional biopsy showed that this was a benign bone lesion, suggesting the provisional diagnosis of an ossifying fibroma, fibrous dyplasis and ameloblastoma.
For accurate diagnosis and proper treatment plan of the lesion the maxillofacial computed tomography (CT) scan was performed to obtain valuable information regarding the size and extent as well as the localization of the pathological patient's condition affected by tumor.
Medical Image processing for 3D modelling and 3D printing Patient underwent the preoperative maxillofacial and pelvis computed tomography (CT) scans (GE Brightspeed Edge 8, USA). For the exams, an image acquisition protocol was used 1 mm for increment between slices and 1 mm of thickness, zero degrees of gantry inclination of the orientation. The CT scan images of the maxillofacial region (257 images) and the pelvis region (585 images) were stored in Digital Imaging and Communications in Medicine (DICOM) file format, which is the international standard format for medical imaging.
To 3D visualize the CT imaging data as in DICOM file, the images were imported into 3D Slicer (version 4.10.1) software, which is a free, open-source software package for 3D visualization of medical images and image analysis. This software provides the advanced 3D visualization functionalities such as surface rendering as well direct volume rendering. The uploaded images were processed, edited and further segmented to perform the accurate 3D visualization of patient's anatomical particularities on the basis of desired tissues and pathophysiology in the form of 3D rendered virtual models through volume rendering in 3D Slicer software.
In general, 3D virtual models generated in 3D Slicer software are surface models that are build up out of triangle meshes. 3D virtual models as saved in STL (STereoLithography or Standard Triangulation Language) file format were further edited and refined using Meshmixer® 3.5 version (Autodesk, Inc., USA) to generate the error-free 3D printable meshes, to remove all artifacts, and to minimize surface imperfections amendable to 3D printing. Meshmixer is also an open-source software for working with triangle meshes.
All the resulting 3D virtual models completed in Meshmixer were saved once again in STL file format and loaded into the Flashprint software (FlashForge Corporation, Zhejiang, China) to set up parameters such as resolution, temperature, wall thickness, support structure, etc. for 3D printing. At the final stage, the resulting STL files were exported to a Dreamer NX 3D Printer (FlashForge Corporation, China) Fused Deposition Modelling (FDM) ultimately to fabricate patient-specific, highly accurate 3D anatomical models as 3D physical models.
The materials used for 3D printing were polylactic acid (PLA), and acrylonitrile butadiene styrene (ABS) filaments from Material 4print, Germany or Polymaker, China. Printing took around 8 to 10 hours for a typical physical model, e.g., patient's skull, however this was variable depending on the size of the model required.

Results
On clinical examination, the patient had a severe swelling on the left side of cheek. There were neither teeth displacement nor teeth root resorption in the lesion area, but it was shown serious malocclusion as well as restrictive mouth opening. Based on the histopathological analysis it was diagnosed a benign bone tumor, suggesting the ossifying fibroma, fibrous dyplasis or ameloblastoma provisionally [17][18][19][20].

3D Visualization of CT scan images and creation of virtual 3D models
Maxillofacial CT scan generated a set of parallel cross-sectional images -also called slice images -in 2D matrix format. To 3D visualize the CT imaging data as in DICOM file, the images were firstly imported into 3D Slicer Software.
Stacking the slice images together and resampling the 2D data in 3D slicer software allowed to form a volume, so-called a volumetric data set, which contains the details of structures hidden inside this volume (See Fig. 1). Once uploaded DICOM files, the images were processed, manipulated and edited to perform the accurate 3D visualization of patient's anatomical particularities on the basis of desired tissues and pathophysiology in the form of 3D virtual models through volume rendering (   After fine tuning in 3D Slicer, the generated 3D virtual models for the face and facial skeleton were imported into Meshmixer software using STL-format to inspect the fine anatomical detail and pathological conditions affected by tumor, which would be usually more difficult to discern in conventional 2D images alone. The resulting 3D rendered virtual models on the computer screen would be interactively inspected, e.g., rotating, tilting, slicing, and extracting region of interest (ROI) etc., to identify the more detailed anatomical and pathological information. The close inspection of the virtual 3D model for the facial skeleton showed a clear anatomical differences (contour differences) between on the left and right of the maxilla and mandible, respectively ( Fig. 4a and 4b). While there was no anatomical defect on the right maxilla and mandible area (Fig. 4a), as clearly seen on the left side (Fig 4b), the lesion presented irregular osseous defects, including a partial rupture of zygomatic arch and also extensive resorption of zygoma and coronoid (Fig. 4c). In addition, an expansive mass of cortical plate on the ramus region in the mandible was identified in this 3D virtual model.

Multiple Surgery
According to the diagnosis with CT scan and histological analysis it was verified a benign tumor with a slow-growing asymptomatic swelling causing facial asymmetry, but if the swelling could progress to considerable size and then the abnormal cell growth easily infiltrates, it may destroy surrounding bony tissues, resulting in loose teeth and malocclusion, which affect a patient's quality of life drastically. Based on these results, it was decided to undertake multiple surgical procedures, including the management of damaged zygomaticomaxillary complex (ZMC) and a segmental mandibulectomy, i.e. a unilateral partial resection of mandible affected by tumor, and ultimately followed by maxillofacial reconstruction. In these instances, we created an appropriate strategy to perform the fewest number of surgical procedures necessary to obtain optimal results in the shortest possible time.
VSP with virtual 3D models on the computer screen Procedure 1: Repair planning for ZMC defect Among the multiple surgical procedures required for restoration of maxillofacial reconstruction, at first we started to manage the damages in ZMC. Dedicated 3D virtual model of facial skeleton (Fig. 4c) clearly demonstrated the complex bony defects in the ZMC resulting from tumor, including an isolated unilateral rupture of the temporal process of the zygoma and a segmental defect on the zygomatic process of the temporal bone, in particular on the zygomatic arch. Shortly, a free-floating rest of the damaged zygomatic arch was retained barely to the temporal bone.
The zygomatic arch provides the facial width because of its bowed convex shape and figures in the lateral contour to the midface. The zygomatic arch is also vital in the mastication system. However, if damaged, defects of zygomatic arch and zygoma give rise to not only aesthetic defects, e.g. flattening of the cheek, lateral midface depression, and asymmetric reduction of facial width, but also functional defects, e.g., restrictive mouth opening, malocclusion etc.
It is important for treatment of defects in ZMC to select the adequate technique and instrument because inadequate stabilization and reduction of zygomatic arch may result in malunion or asymetry [25]. Management includes repositioning and fixation of the damaged bones to restore normal facial appearance, such as cheek prominence and craniofacial symmetries, and also masticatory function.
After the careful evaluation of the residual maxillary bone, a facial implant was designed virtually to restore the facial contour and to balance the zygomatic bone as cheekbone with the rest of the face. This facial implant was called hereafter a zygomatic implant. At first it was defined the resection line for osteotomy of bony defects on ZMC. The damaged areas with sufficient margin in the ZMC were resected parallel to the zygomatic arch, where eventually will be used the basement for placing the implant (Fig. 5).
In order to design the zygomatic implant, the resection line on the site of the defect was mirrored in the healthy side of the right zygomatic arch, which in turn was segmented with the tool of plane cut [21]. The corresponding segmented mirror image was reflected into the defect side along the mirror axis and used as template to complete the implant contour design. After cutting off the non-overlapping area of the mirror image the resulting image was converted into a 3D digital model, by which the zygomatic implant was created in a form of the 3D virtual model ( Fig. 6 a, b, c). Finally, the resulting virtual zygomatic implant, in which three holes were predesigned for implant fixation on the resected ZMC during the operation, was secured in place with 3 titanium screws of 2mm diameter (Fig. 6d).
Zygomatic Implant functions just like a real bone, which looks and feels natural, and integrates seamlessly with the structures of the face. As a consequence by placing a zygomatic implant on the cheekbones it provides the patient with an exceptional anatomical fit and enhanced aesthetic results that has a big impact on the patient's quality of life after real surgery.
In general, such zygomatic implant can be made of a variety of materials depending on the anatomical location and surgical preference. The most common material is solid silicone. It has been used for many years because it is safe, reliable and durable. It is also low cost and can retain extremely fine detail.

Procedure 2: Repair Planning for Mandible
The primary resection of the afflicted part of the mandible was carried out with virtual 3D models derived from the preoperative CT-data (Fig. 3). To more accurately diagnose the pathologic conditions on the left mandible, the ROI was cropped and then closely inspected with direct volume rendering technique. (Fig. 7). Radiographically, the tumor area showed that there were small localized multiocular, mixed radiolucent radiopaque lesions, exhibiting a perforation of cortical plate at the coronoid area (arrow 1) and an expansion of cortical plate on the ramus area (arrow 2and 3) with varying size and feature (Fig. 7a). In addition, bone resorption was clearly observed in the coronoid process of the left mandible compared with the right mandible.
Because of the large size of the tumor-affected area in mandible, it was planned a segmental resection of the mandible (mandibulectomy). To remove the tumor-affected area in the ramus completely, at first, the precise resection lines were acquired (Fig. 8). The tumor defect areas were clearly defined by the extension and dimension of the pathological defects closely inspected on the virtual 3D model and then outlined with resection lines on the ramus by which it could be precisely separated into the defective and healthy parts. These resection lines can assist the surgeon to accurately determine the position and direction for cutting of the mandible during the real surgery. Hereby keeping in mind, the tumor defect areas should be completely excised with sufficient wide tumor-clear margin where possible (in our case at least 10 mm from the tumor) [27].

Mandible Cutting Guide
For segmental mandibulectomy a life-size 3D virtual model of mandible as a template was generated by segmentation of the virtual 3D virtual model of Fig.8. To assist surgeon to perform mandibulectomy easily, conveniently and accurately, a cutting guide was designed with the Boolean operations in the Meshmixer software, according to the well-defined resection lines (Fig. 9).
The cutting guide was used to control the position and direction of osteotomy for complete tumor resection.
The site of cutting guide to be fixed to the mandible exhibited the identical surface topology of the mandible that is a prerequisite for accurate positioning the cutting guide on the mandible. The surface of the cutting guide was marked with arrows, where the arrows on the upper surface indicate its exact position on the ramus to be placed -between both sides of coronoid process, whereas a double sided arrow indicates where the cutting to be carried out in the surgery. Once the cutting guide was adapted at an exact place on the mandible and then secured with 2 cortical locking screws on the posterior border of the mandibular ramus, the segmental mandibulectomy was performed using a sagittal saw along the proximal and distal edges of the mandible cutting guide (Fig. 10). In this way, the planned mandibular resection can be accurately performed at surgery.
A careful evaluation on the 3D virtual models clearly indicated that the condyle remained intact (Fig. 10a). In general, accurate repositioning of the condylar head in the temporomandibular fossa is difficult to maintain the maxillomandibular relationship. Because false condyle position and condylar disk displacement often induce temporomandibular joint dysfunction associated with malocclusion [28,29]. Therefore, it was decided that the intact mandibular condyle should be preserved in its initial position without further effort.
The defect area was resected with a mandible cutting guide and then replaced by healthy contralateral autologous bone graft to create the repair model. To this end, the resected part from the mandible will be used as template for harvesting of graft segment from a donor site (Fig. 12c).

Selection of donor site
Since there is a wide variety of donor sites for selecting and harvesting the type of the bone graft with the sufficient amount volume and shape have to be decided by the surgical team after intensive discussion. Bony reconstruction of the discontinuity in mandible followed by mandibulectomy can be achieved by different types of autologous bone grafts. There are several different opinions about the best possible autogenous bone graft to use in the maxillofacial reconstruction.
Currently autogenous bone grafting is the gold standard, which is reliable and highly successful with excellent long-term functional and aesthetic outcomes. The most common autogenous bone grafts are derived from the fibula, iliac crest, scapula, calvarium, rib or radius [29][30][31].
The selection of donor site for autogenous bone graft depends mainly on the defect size, location, need for soft tissue and status of the recipient vessels. In addition, the availability of donor sites, e.g., adequate amount of bone volume, is also important [33]. To decide which bone graft is best to restore bony continuity and facial contour, the setting at the defect site is crucial.
Through the evaluation according to the size of resected bone, including length and volume, the defect size was relatively large in dimension with 3 x 4 x 1 cm 3 (height, length and width respectively) but no additional tissue was needed. Therefore, for achieving sufficient anatomical bulk being replaced the continuity defect in mandible we decided to harvest the bone graft from patient's pelvis. Because of its availability of large amount of bone with a compact cortex and rich cancellous blood supply in combination with the lower donor site morbidity compared with other types, the iliac crest in pelvis can provide an excellent intact facial function and aesthetic outcome for the mandibular reconstruction of segmental defects.

CT Pelvis
After decision of the donor site, the patient's pelvis was scanned by CT. And then CT images were processed to create the 3D virtual models of the iliac crest bone on the computer screen to get precise information on the possible shape, size and placement of the bone graft needed (Fig. 11). Based on the detailed evaluation the location and size of the bone graft to be taken from the right-sided iliac crest should be carefully determined with precision.

Patient-specific surgical tools
With the resected part from mandible as a template for harvesting bone graft, the graft shape and cutline were defined on the virtual 3D model of the segmented right-sided iliac crest (Fig.  12 a) accurately, as is the guide for cutting and shaping primarily.
To precisely define the area for harvesting the bone graft on the ilium two surgical tools were patient-specifically designed which were fitted uniquely on the donor site in only the planned position with defined orientation (Fig. 12b). The one was an iliac crest cutting guide to harvest the bone graft indicated the outline of the planned resection area (Fig. 12c). Such cutting guide therefore included all necessary information, such as graft size, angulation, and osteotomies, according to the virtual plan. The other was a jig which was designed and used to assist the iliac crest cutting guide for locating at the exact positon on pelvis.
Ultimately, the iliac crest bone graft was resected virtually as guided by the iliac crest resection guide, then harvested, as shown in Fig. 12e. Using well-defined harvesting strategy with the uniquely fitted patient-specific surgical tools, e.g., an iliac crest cutting guide and jig, one can avoid harvesting bone unnecessarily, by which may be minimized postoperative donor site morbidity and also reduced the harvesting time. Once the iliac crest bone graft was released, that should be shaped precisely to match with the discontinuity caused by segmental defect in the mandible (Fig. 13). If reassembled all the parts together in planed position with right angulation by which it was achieved an excellent virtual mandible symmetry as shown in Fig. 13c, the remained mandibular stump, iliac crest bone graft and condyle were bridged together with osteosythesis plates and fixed by osteosyntesis screws for the best possible restoration of the patient's appearance and oral function.

Shaping and fixing
According to the stress states on the mandible during mastication the number, orientation and length of the osteosythesis plates were determined with well-defined precision: the one was at the area where the tensile stress was dominated, on the other hand the other was at the location where the compression was mostly subjected (Fig. 14a) [34,35]. For the best possible 3D adaptation of the osteosynthesis plates throughout the rest of mandible stump, the bone graft and the preserved condyle the osteosynthesis plates were pre-bent whose curvature and contour were adjusted preoperatively on the virtual 3D mandible model and trimmed for a customized fit ready for the real surgery (Fig. 14b). Thus, the patient-specifically pre-bent osteosythesis plates exhibited an anatomical shape precisely fit to the maxillofacial contour without further bending or adjustment at the time of surgery, so that the surgical time will be reduced significantly.
As a final stage, surgeon then directs virtual reconstruction by superimposing the patient's own 3D reconstructed iliac crest bone graft onto the mandibular defect to recreate the native mandibular contour through a trial-and-error process, optimizing the number and cutting plane of the osteotomies, bone-to-bone contact, and segment lengths.

Pre-surgical simulation with 3D printed physical models
Once VSP with 3D virtual models was accomplished, the relevant anatomical models of the regions of interest being involved in the real surgery were then printed out in 3D, under the following aspects; firstly to improve and complete the virtual surgical planning in the feasibility aspect, secondly to simulate a real surgery in the accuracy and effectivity aspect, lastly to prepare the surgical tools and implant in the technological aspect. With these 3D printed physical models (Fig. 15), on anywhere without presenting the patient surgeons can simply and precisely recognize the lesion affected by tumor in a 3D environment, in order to discuss the treatment procedure within a surgical team, for designing a more precise surgical procedure.
Accurately 3D printed anatomical models were used as surgical templates to precisely simulate for osteotomy both mandible and ZMC. Firstly, to simulate the management of defects on ZMC we fabricated the zygomatic implant by 3D printer, hereby used the STL file created in Fig. 6C. The zygomatic implant was instantaneously used as a cutting guide for accurate osteotomy of heathy ZMC. The prepared basement for repositioning the zygomatic implant is shown in Fig 16a. The screw hole locations on the zygomatic implant were used to position the implant at the appropriate location exactly (Fig. 16b). Fixation was performed with three osteosynthesis screws through the existing holes on the prepared ZMC.
Such a customized zygomatic implant allows us to precisely restore the ideal facial contour and symmetry. As the next, to simulate mandibulectomy we produced a cutting guide, which was uniquely fitted on the 3D printed mandible (Fig. 17a, b). At first, the cutting guide was fixed temporarily to the lateral side of mandible using two osteosynthesis screws through the existing holes, and then the defect area was resected using an oscillating saw, along the both sides of cutting guide as marked with arrows, as shown in Fig. 17c. After the osteotomy, the remaining mandible, shown in Fig. 17d, should be linked and repaired with a bone graft. For harvesting the autogenous bone graft from iliac crest, the iliac crest of the part pelvis was 3D printed in the real size as used a simulation template. To accurately position of a cutting guide at the exact location on the iliac crest one additional jig was designed and manufactured by means of 3D printing (Fig. 18a). As a consequential procedure, the cutting guide combined with jig together was fixed onto the iliac crest and then removed the jig by keeping the cutting guide position at the defined position, so that one can precisely resect the iliac crest to harvest the bone graft with exact volume without having to wait for the exact measurements (Fig. 18).
The harvested bone graft was refined and reshaped in vitro with high accuracy according to the virtual models as in Fig. 13, to adapt to the discontinuity in the mandible. The resulting bone graft was fixed to the appropriate position on the mandible using 2 osteosynthesis plates with 9 titanium screws at the exactly defined position (Fig. 19). Hereby keep in mind, any aberration in mandibular structural alignment may lead to functional disturbances due to malocclusion. Furthermore, adjusting and cutting the osteosynthesis plates to precisely fit the mandibular contour during the course of the operation was very difficult and even time consuming. Therefore, prior to the actual surgery precisely adapting the reconstruction plates to the mandibular contour is prerequisite.
The customized osteosynthesis plates were primarily adjusted to the predefined orientation using the contour line as designed in the virtual planning (Fig. 14b) and then pre-bent to bridge between the residual mandible stump, iliac bone graft and condyle, for achieving the best possible restoration of the patient's appearance and oral function.
Accurately 3D printed anatomical model of the resected mandible and zygomatic bone were used as surgical templates to precisely replace the maxillofacial bone defects both with iliac crest bone graft and zygomatic implant by adapting the pre-bent osteosynthesis plates with specific length of screws.
Thus, simulation with 3D printed anatomical models for maxillofacial reconstruction results in an improved clinical outcome in both aspects of function and aesthetics with the potential for improved quality of life for the patient.
After adjusting and re-validating in a pre-surgical simulation with 3D physical models by doing repetitive practice, the entire surgical procedure will be perfectualized and fully understood by surgical teams as a training platform in a non-threatening environment before they practice on patients (Fig. 20). Finally, the expected surgical outcome as in a hand-on 3D printed model was suggested virtually by reconstructing maxillofacial contour symmetry (Fig. 21). It is worth noting here that this models can be shown to the patient for his decision making to accept this kind of surgical procedure. Through the simulation with 3D printed physical models the feasibility and validity of predesigned plan can be verified in the clinic prior to real surgery.

Discussion
With advances in digital imaging and computer science, 3D visualization of medical images has become a routine practice in medicine not only to diagnose diseases, but also to prepare therapeutic planning. [36] Volume Rendering 3D visualizing the volumetric data set from CT and MRI scan is called volume rendering that makes possible effectively to extract meaningful information on 3D structures from the volumetric image data set and to display them on a computer screen [22,37]. The results in 3D images displayed on screen allow the radiologists to visually explore, analyze and effectively interpret them for the evaluation of suspected disease, as well as the surgeon more accurately to prepare surgical planning in accordance with the anatomical relationship of one structure to another. [38] There are a number of different techniques for volume rendering, but most can be commonly divided into the following two classes: surface rendering and direct volume rendering (DVR) [39].
Surface Rendering [40], also known as thresholding-or surface-based (binary) techniques, is the extraction and 3D visualization of the apparent surface of interest (iso-surface) within a volumetric data set. The iso-surface, which is formed by a set of data points with equal value or density within a volume, is usually approximated by a set of polygonal meshes (generally triangles in 3D printing) that form the surface of an object. Surface-rendered image provides a realistic 3D relationship of the surface of a structure of interest, which is particularly appropriate for studying fine details such as bone surfaces. This technique, however, fails to display the details underneath the surface, such as subcortical pathology.
In contrast, direct volume rendering (DVR) [41,42], which is also known as percentage-or semitransparent volume-based (continuous) technique, incorporates the entire volumetric data set into a 3D image; without the use of an intermediate geometric representation for iso-surfaces.
With direct volume rendering, one can effectively visualize the details beneath the surface, e.g., in particular resolving hidden areas of interest -subcortical lesions, which may contribute to a more comprehensive understanding of pathologic conditions and processes. As a consequence, DVR greatly increases the sensitivity for localizing smaller lesions and thus improving the overall accuracy for localizing lesions, which is essential for tumor resection and treatment plan [43].

3D virtual models
Recent advances in computer graphics for medical imaging make interactive volume visualization both in virtual and physical modes possible. [44] Using specialized software, here in this work 3D Slicer, the acquired medical imaging dataset can be rendered into 3D virtual models of the patient. With the 3D virtual models, which are digital 3D renderings of the regions of interest from the patients, all structural details hidden in the volumetric data set can be displayed in an appropriate 3D context on the computer screen interactively [45].
Interactive means that the 3D virtual models can be electronically manipulated, segmented, cropped, rotated into different view angles at any depth within the volume, and zoomed in and out during the examination to provide complex spatial relationships in a 3D environment.
3D virtual modeling can give the surgeon a better understanding of the anatomy and pathology to assess most of the surgical difficulties and risks, by which a clear idea creates about how to optimize the surgical approach before actual surgery takes place.
In this work, we generated a precise preoperative surgical planning to perform osteotomies resections of the recipient and donor sites for harvesting and placing the bone graft as well as to design a customized implant for damaged zygomatic arch and surgical tools such as cutting guides and jig virtually.

VSP
Such a well-defined preoperative surgical planning method is so-called Virtual Surgical Planning (VSP) that involves the visualization of each step of the surgical procedure in a virtual environment using 3D imaging computer software [46]. Ultimately, the resulting VSP can be precisely applied to the patient in the operating room that means transferring the virtual planning to real time surgery.
Using a combination of 3D virtual modelling and VSP it is possible to increase the accuracy of the surgical procedure significantly, -to decrease the necessity for intraoperative trial and error drastically, -to design the customized implant and surgical tools precisely, thereby to improve the predictability of surgical outcomes effectively [47], and -to reduce the overall duration of surgery efficiently. All the above mentioned points are beneficial both patients and surgeons. VSP helps to inform patients and train surgeons by giving them a means to visualize exactly what the surgeon will do as well as how the procedure will work in the actual surgery. As a consequence, VSP can provide more security and reliability to the process as a whole.
Although the virtual 3D models can provide the surgeon with interactive 3D view of the detailed anatomy of the specific regions of interest, they are strictly confined on the computer screen, so their effects on the preparation of a successful surgical plan are strictly limited. In addition, during medical consultations most surgeons often use 2D images, in occasional cases show even 3D volume rendered images, which are also displayed on a flat 2D computer screen, however many patients may not fully conceptualize these images and also not clearly understand the clinician's interpretation of the true structures of the underlying anatomy.
As compared to 3D virtual models, the physical models generated by 3D printing allow surgeons to hold in their hands, touch, feel, and visualize all aspects of the anatomy and pathology in any different view angles.
3D physical models 3D printed models help to overcome the limitations of consultations performed with 2D images, due to superiorly encapsulating anatomic spatial relationships in both visible and tangible manners. That way surgeons can use the 3D physical models through hands-on interaction to communicate the surgical steps to the clinical team and then to explain an upcoming surgical procedures to patients and their families without being tied to a computer monitor in advance of the actual treatments.
Accurate and touchable 3D physical models would allow any trainees, such as novices, residents and medical students to learn the surgery with the basic operative steps of specific procedures, enabling effective training in a safe environment to significantly improve their knowledge and skills [48]. Furthermore, in the complicated cases, where the patients undergoing new, uncommon or complex multiple surgeries, 3D physical models are inevitable and beneficial with, by which it increases surgeon's self-confidence approaching difficult anatomies competently with new medical devices, which would be better accommodated the patient's anatomy.
The individual variances and complexities of each patient make the use of 3D-printed models ideal for surgical preparation on a case-by-case basis, particularly in cases where the anatomy is complex and there are numerous critical anatomical structures in close proximity to the operating field. As a result, surgeries will be able to be customized, optimized and executed in shorter time, lesser risks to the patient and more predictable surgical outcomes, thereby significantly improving the quality and efficiency of the surgery.
In practical point of view, the customized cutting guides uniquely fitted to the mandible and pelvis were designed with virtual 3D models and then manufactured by 3D printing with an adequate precision range for using them in the real time surgery. It is worth noting that there is a correlation between the amount of bone harvested and postoperative complications; therefore, the harvested bone graft should not contain unnecessary bone [49]. It means that as much as possible to cut off the bone defect but as much as possible to reserve the healthy part. Through the evaluation of the size and volume of the resected mandible bone segment the patient's pelvis was decided to get an exact amount of bone for optimal placement in the resected mandible [50].
The intraoperative shaping procedure of the iliac crest bone graft for precise positioning onto the resected mandible is a hard task even for the experienced surgeon and consequently increases the ischemic time, which is one of the main causes of transplant failure. Therefore, the shaping procedure during the operation should be as quick as possible. Furthermore, intraoperative bending and cutting of the reconstruction plates as well as adjusting the screws with appropriate lengths, all of which definitely depend on the anatomical complexity of resection of the recipient site and surgeon's skill, may not be accurate enough and could significantly prolong the operation time. Therefore, without any intraoperative measuring it is necessary to prepare the patient-specific reconstruction plates and screws with the appropriate lengths prior to surgery.
Traditionally, reconstructive surgery for complex facial deformities was heavily relied on the surgeon's performance, proficiency and competence achieved from earlier experience. However, with the help of precise virtual surgical planning and unique cutting guides it is allowed nowadays to prepare the recipient and donor site precisely, to avoid harvesting bone unnecessarily, to reshape the bone graft adequately, and thus to be greatly shortened the duration of the surgery in the operating room.
To achieve the best possible facial contour in terms of esthetics and functional recovery, which are the ultimate goal of maxillofacial reconstruction and thus considered as a measure of evaluation for the clinical outcomes, the need for precise virtual surgical planning and presurgical simulation with the life-size 3D printed models cannot be over emphasized.

Pre-surgical simulation
In order to verify the feasibility and validity of predesigned virtual surgical plan prior to real surgery, we performed a pre-surgical simulation with the life-size 3D printed anatomical models [51]. The life-size 3D printed anatomical models were used to simulate many kinds of surgical procedures that allowed to identify the best promising surgical strategy for the patient in terms of results and timing.
The 3D-printed anatomical models of the regions of interest, especially for a patient showing the sophisticated anatomy structures involved in the surgery, are valuable aids for preoperative planning and pre-surgical simulation. During in vitro simulation with 3D physical models, the surgeon can foresee intra-operative complications in both anatomical and technical aspects of the procedure, e.g., how to handle the patient's anatomy for accessing the tumor and performing the osteotomy, which may result in improved intraoperative precision with the reduced operating time. Furthermore, surgical simulations involving the life-size 3D physical models allow trainees, such as trainee surgeons, residents and medical students, to practice with a diverse range of scenarios repeatedly as many as possible without putting a patient at risk [52,53]. As a consequence, a pre-surgical simulation is considered to be an excellent way to learn the real surgery in a low-risk environment. Pre-surgical simulation enables the surgeon to improve the surgical skills, avoiding accidental injury, eliminating the risk of surgical errors, and thus improve self-confidence during actual surgery significantly.
Once the surgeon accumulates sufficient experience with the pre-surgical simulation and rehearsal, without thinking about the surgical plan the predesigned surgical plan can be simply executed in the operation room, which guarantees the quality and accuracy of surgery and thus potentially increases the success probability of the real surgery.

Conclusions
3D technologies, e.g., 3D rendering / modelling and 3D printing, were applied to analyze complex anatomies, optimize the surgical repair or intervention planning and guide surgical strategy for maxillofacial reconstruction caused by tumor. Based on the 3D volume rendering for CT images the 3D virtual models for the patient-specific pathological anatomy were generated for a virtual surgical planning, hereby to manage almost every scenario that could arise during the real-time surgical procedure as well as to design the customized implant and intraoperative surgical tools with greater accuracy. Prior to entering the operating room the surgeons can simulate the surgery assisted by 3D printed patient-specific anatomical models as templates with customized implant and autogenous bone graft. The combination of more accurate virtual surgical planning with 3D virtual models and pre-operative simulation with 3D physical models can lead to optimize the surgical intervention for complex maxillofacial defects affected by tumor in efficient manner that in turn results in the more predictable surgical outcomes and reduction in the duration of surgery and surgical risks significantly.
Further studies should address the transfer of the surgical strategy proposed in this work to the real-time surgery for extreme complex maxillofacial reconstruction.