Facial Reconstruction: A Systematic Review of Current Image Acquisition and Processing Techniques

Introduction: Plastic and reconstructive surgery is based on a culmination of technological advances, diverse techniques, creative adaptations and strategic planning. 3D imaging is a modality that encompasses several of these criteria while encouraging the others. Imaging techniques used in facial imaging come in many different modalities and sub-modalities which is imperative for such a complex area of the body; there is a clear clinical need for hyper-specialized practice. However, with this complexity comes variability and thus there will always be an element of bias in the choices made for imaging techniques. Aims and Objectives: The aim of this review is to systematically analyse the imaging techniques used in facial reconstruction and produce a comprehensive summary and comparison of imaging techniques currently available, including both traditional and novel methods. Methods: The systematic search was performed on EMBASE, PubMed, Scopus, Web of Science and Cochrane reviews using keywords such as “image technique/acquisition/processing,” “3-Dimensional,” “Facial,” and “Reconstruction.” The PRISMA guidelines were used to carry out the systematic review. Studies were then subsequently collected and collated; followed by a screening and exclusion process with a final full-text review for further clarification in regard to the selection criteria. A risk of bias assessment was also carried out on each study systematically using the respective tool in relation to the study in question. Results: From the initial 6,147 studies, 75 were deemed to fulfill all selection criteria and selected for meta-analysis. The majority of papers involved the use of computer tomography, though the use of magnetic resonance and handheld scanners using sonography have become more common in the field. The studies ranged in patient population, clinical indication. Seminal papers were highlighted within the group of papers for further analysis. Conclusions: There are clearly many factors that affect the choice of image acquisition techniques and their potential at being ideal for a given role. Ultimately the surgical team's choice will guide much of the decision, but it is crucial to be aware of not just the diagnostic ability of such modalities, but their treatment possibilities as well.


INTRODUCTION
The advancement of modern imaging techniques in plastic and reconstructive surgery has led to ever improving patient outcomes, allowing the surgeon to visualize and gain a better spatial appreciation during complex surgical procedures. Of those utilized, three-dimensional (3D) imaging has become commonplace due to its accuracy, precision, and versatility (1). 3D imaging has multiple modalities, each with their own advantages and disadvantages; these include, but are not limited to, magnetic resonance imaging (MRI), computed tomography (CT), 3D photography and handheld scanners using sonography ( Table 1). All these modalities and their respective sub-modalities play their own role in the diagnosis, planning, and treatment of the patient, thereby, becoming a cornerstone of preoperative, perioperative, and postoperative care (2). In facial reconstructive surgery, the use of detailed facial imaging techniques is imperative for such an anatomically intricate area of the body; for this reason, there is a clear clinical need for hyperspecialized practice (3).
The advancement of facial reconstructive surgery has coincided with the evolution of the imaging modalities available for patient assessment, surgical planning, and postoperative follow-up. For decades, prior to the development of CT and MRI in the late 1970s, clinical 2D photography was utilized as the primary imaging modality for the objective measure of surgical outcomes (4). The use of 2D clinical photography continues to be a cornerstone of plastic surgery due to its availability, ease of obtainment, and reproducibility when re-examination is required (5).
The introduction of CT and MRI scans in the early 1980s as well as recent advancements in 3D photography and ultrasound scanners permit detailed geometric, topographical, and volumetric analysis, allowing the surgeon to form a more accurate clinical picture for preoperative surgical planning and/or intraoperative surgical guidance (6). Advancements in imaging modalities and increasing literature in this field necessitates a comprehensive review of the advantages and disadvantages of each imaging technique to act as a guide for clinicians involved in facial reconstruction. In addition to the high resolution of these imaging modalities, the ability to combine these with angiography components adds an extra and indispensable element to the preoperative planning process (7,8).
With the variety of imaging techniques and computer processing systems available, identifying which case-specific imaging modality to use has become a challenge, with each modality having a situational-dependent advantage. The importance of an evidence-based imaging choice cannot be overstated when picking a modality; it becomes easy for bias to be formed through familiarity rather than best practice. Despite this, there is limited literature available that strives to provide conclusions and therefore limited guidance for surgeons on which imaging modality to utilize for specific facial reconstructive procedures.
Surgical planning is a broad term that encompasses several meanings; here, it refers specifically to the use of image acquisition and processing techniques used in the context of facial reconstructive plastic surgery. The aim of surgical planning is to simply acquire the most comprehensive representation of the structure being operating on, thus, visualizing the surgery about to be undertaken. This visualization can be done with single or multiple imaging techniques, often multimodal approaches providing the most comprehensive picture. Surgical planning begins with image capture of the specific anatomical area, these images are then rendered using modality-dependent processing techniques (9). The result of this is a virtual 3D model that allows for visualization of the anatomy from different angles and provides the surgeon with a "mental map" that allows for surgical navigation throughout the operation (10). With these tools at their fingertips, a clear stepwise approach can be taken for each operation. In addition to this, these walkthroughs can be shown to patients to allow education and inform the consent process and therefore improve patient outcomes (11).
The aim of this systematic review was to highlight all the available image acquisition and processing techniques that have been utilized for bony facial reconstruction in the literature. The goal is that this forms a comprehensive summary of the available bony facial imaging techniques for surgeons involved in facial reconstruction and guides decision making based on patient-specific needs.

Search Strategy
A systematic review protocol was developed in accordance with the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Figure 1) (12) to evaluate the types of imaging techniques and their clinical variability in the literature To identify all relevant papers, a comprehensive search strategy was developed and pretested, with an example of the search strategy used; the full search having being provided later in the Methods section. Searches were performed in Web of Science (Web of Science  Core Collection, BIOSIS Citation index, KCI-Korean Journal  Database, MEDLINE, SciELO Citation Index), PubMed, Scopus, OVID, and Embase. The review was additionally registered on the International prospective register of systematic reviews (PROSPERO) (13).
The specific search and selection strategy performed is highlighted below and is based on the initial search strategy registered with PROSPERO. The keywords and strategy utilized is as follows; "(((facial OR maxillofacial) AND (reconstruction OR recon surgery)) AND ((image technique) OR (image acquisition) OR (image processing) OR ((computed tomography) OR (magnetic resonance imaging) OR (ultrasonography) OR (3D photography))))" This search strategy provided a clear but precise identification of relevant articles in the literature. These papers were then screened further by the use of specific eligibility and exclusion criteria.

Eligibility Criteria
Eligibility for article inclusion were (1) all studies included involved facial imaging techniques; (2) bone reconstruction; (3) included only imaging modalities aiming to visualize bony structures or imaging modalities aiming to visualize facial reconstruction; (4) clinical patient outcomes; (5) English language articles only.

Exclusion Criteria
Studies were excluded if they met the following criteria: (1) no clinical component; (2) involved non-facial surgical techniques; (3) purely soft tissue reconstruction; (4) contained no surgical component; (5) were not available for viewing. Review articles and commentaries were also excluded.

Study Selection
Two reviewers (S. P. T. and M. E. S.) independently reviewed the studies with differences resolved by a third reviewer (R. S. A.). Titles were initially screened to exclude duplicates and further screened using the abstracts against inclusion and exclusion criteria. Finally, full text review of the remaining articles was performed to assess for eligibility.

Data Extraction and Main Outcomes
Data were extracted from the selected studies using a standardized format (Microsoft Office Excel 2019). The initial tabulated data collection included: anatomy, patient demographic, study demographic, methodology, study design, primary and secondary outcomes, complications, and clinical availability.

Risk-of-Bias Assessment
Studies were examined for bias using the Cochrane tool (RoB 2.0) (14) as a guide alongside modified versions of the tool and evaluations of the tool itself (15). The literature was separated into cohort-based ( Table 2) and trial-based ( Table 3) categories due to the differing criteria needed for assessing the bias (16).

RESULTS
The initial study numbers that were found (after duplicate removal) were 6,147, which led to the final inclusion of 75. Papers were removed from the screening and data extraction stage if they did not abide by the criteria laid out in the PROSPERO protocol. The papers were categorized into the anatomical area of the face that the work correlated to and then arranged by the date of publication ( Table 4).
The frequency of the studies was looked at depending on the year of publication ( Figure 2) and it was found that there was an almost exponential rise after 2011 which would be explained by the readiness of the technology available and the emergence of 3d printing abilities.
Risk-of-bias assessments were carried out using different tools that depended upon the type of study that was being assessed, i.e., cohort vs. trial-based study.

DISCUSSION
As can be seen from the analysis and results collation, there are many varying papers in the literature that use facial imaging in the clinical environment. Whether for diagnosis, treatment or prognosis, these techniques are commonplace in surgical theaters throughout the world and will become even more so in the future. It becomes a difficult task to analyze all the studies in a quantifiable sense due to the sheer variability between them, but it would be prudent to highlight what could be seminal papers in the field so that they can be turned to as reference. Not only are the key points of the papers crucial, but the impact factor of the journal and the ideals that they hope to achieve in future studies.
The seminal papers for the imaging modalities discussed have been selected through multiple means following a rigorous search strategy and certain criteria. These include but are not limited to: the papers impact factor, cohort study size, study type and low bias. The seminality of the papers selected has been based on the authors being the first to report specific imaging modality technique or being the first to utilize it in clinical practice with reportable and significant current and or future clinical impact.
The limitations of the search and the analysis of data should be noted. With such a large volume of papers to analyze there will be certain branches of facial surgery that were excluded. The foremost being orthognathic surgery. It was considered to include this in the search and analysis of papers, but it was decided that due to the complexity and the vastness of the topic (as it stetches into the world of maxillofacial and dental medicine) to not include these papers. Though one cannot argue the benefit for including these papers in the global scheme of analysis of facial imaging techniques, but with the search that was used, this would not be thorough or robust enough to include the seminal orthognathic papers (87). Similar to this, the use of alloplastic materials are a potential limitation of the review and search strategy. Though the systematic review did not intend to specifically mention imaging of alloplastic implants, we are aware of this challenge and the clinical benefit of the techniques used (88). However, further research would be needed to explicitly look at these criteria. Another limitation of this study is that though a comparable analysis of papers was attempted to be made with data extraction, due to the heterogeneity of the data and the techniques themselves, a true meta-analysis could not be carried out. However, the variability of clinical cases shows that perhaps a meta-analysis would not be of great benefit as it will not provide data that would be translatable to practice.
A detailed and comprehensive research strategy allowed for the exclusion of many papers. Where impact factor was low or cohort size restricted to minimal patients, there was a greater likelihood of exclusion. Although, it should be noted that low impact factor or cohort size did not guarantee exclusion if, in fact,    that paper was deemed "seminal" in nature due to the technique utilized or significance of the study.
Computed Tomography Scanning Systems 1. Wilde et al. Intraoperative imaging with a 3D C-arm system after zygomatico-orbital complex fracture reduction (81) Impact Factor 5.38.

Intraoperative 3D CT C-Arm System
The purpose was to investigate whether intraoperative 3D C-arm imaging is effective technique for assessing the adequacy of fracture reduction in the management of uncomplicated ZMC fractures. They found that the use of 3D CT to assess fracture reduction intraoperatively shows to be an effective tool for evaluating ZMC fracture reduction. Therefore, avoiding additional procedures or need for further imaging.     Bias limitations: Moderate risk of bias from the measurement of interventions (retrospective review).

Ayoub et al. Evaluation of computer-assisted mandibular
reconstruction with vascularized iliac crest bone graft compared to conventional surgery: a randomized prospective clinical trial (44) Impact factor 5.29.

Computer Assisted vs. Conventional Surgery
Virtual surgical planning was based on preoperative CT-data using specific surgical planning software. A rapid prototyping guide transferred the virtual surgery plan to the operation site. The purpose/aim of this study was to evaluate the benefits of computer-assisted mandibular reconstruction with iliac crest bone grafts. It compared the following areas:  This is the first randomized prospective study comparing computer-assisted mandibular reconstruction with conventional surgery using iliac crest bone grafts. The study shows that computer-assisted reconstruction reduces ischemic time and transplant-shaping time at the defect site as well as requiring a smaller amount of harvested bone than conventional surgery. Moreover, a significantly smaller alteration in condyle position could be shown in the computerassisted group.
Bias limitations: High risk of bias from the blinding technique (trial format study).

Heiland et al. Intraoperative imaging of zygomaticomaxillary
complex fractures using a 3D C-arm system (74) Impact factor 3.69.

Intraoperative 3D CT C-Arm System
Purpose/aim of the study was to investigate in a first series patients with ZMC fractures, the practicability and value of intraoperative imaging using the 3D CT C-arm (SIREMOBIL Iso-C3D) after open reduction. They showed that this interoperative imaging technique provided suitable visualization of the facial skeleton postoperatively. This study was done in 2004; therefore, image processing took longer to generate and would not be a technical limitation if carried out now. This paper has similar ideals to Wilde et al. but is important to mention as it paved the way for much of the current modern work.
Bias limitations: Moderate risk of bias from the measurement of interventions (retrospective review).

Zhou et al. Accurate reconstruction of discontinuous
mandible using a reverse engineering/computer-aided design/rapid prototyping technique: a preliminary clinical study (41) Impact factor 5.09.

Reverse Engineering (RE), Computer-Aided Design (CAD), and Rapid Prototyping (RP)
The aim/purpose of the study was to improve surgical outcomes in discontinuous mandibular defects by fabricating patientspecific customized titanium mandibular trays. The was done by utilizing reverse engineering (RE), computer-aided design (CAD) and rapid prototyping (RP) techniques by firstly obtaining a virtual 3D model via spiral CT scanning -the opposite side of the mandible was mirrored to cover the defect area to restore excellent facial symmetry. They then used autogenous bone grafting to restore the bony continuity for occlusion rehabilitation. The trays fabricated using this technique fitted well in all six patients. The reconstructive procedures were easy and time saving. Satisfactory facial symmetry was restored. No severe complications occurred in the five patients without occlusion rehabilitation during a mean 50-month follow-up period. The reconstruction in the patient with occlusion lasted for only 1 year and failed eventually because of bone resorption and infection.
Satisfactory aesthetic results were achieved. However, the rigidity of the cast tray could cause severe stress shielding to the grafts, which could lead to disuse atrophy. A promising technique but some modification is needed for functional reconstruction.
Bias limitations: Moderate risk of bias from the selection of participants (retrospective review).

Shaye et al. Use of intraoperative computed tomography for maxillofacial reconstructive surgery (60)
Impact factor 3.67.

Intraoperative 3D CT C-Arm System
Aim/purpose of the study is to evaluate the time needed to perform intraoperative CT scans during maxillofacial surgery.
In addition, they looked to see if there were any trends toward shorter total scan times as experience is gained with the technique. They also looked to identify the characteristics of cases that required intraoperative revision based on the results of intraoperative CT scanning.

Conclusions
• Current intraoperative CT scanning techniques are rapid, averaging 14.5 min per case (in 38 cases).
• No decrease in total scan time was noted during the study; however, the surgeon most experienced with the CT software had the shortest total scan times.
• Intraoperative revisions were most common in complex cases.
Recommendation from the study is that surgeons consider the use of intraoperative CT imaging for maxillofacial reconstruction, particularly in complex procedures. Bias limitations: High risk of bias from the measurement of outcomes and selection of reported results (retrospective review).

Virtual Planning vs. Conventional Surgery
The aim/purpose of this study was to evaluate the accuracy of mandibular reconstruction and assessed clinical outcomes in both virtual planning and conventional surgery patients. The main outcomes measured: • Operative time • Ischemia time • Postoperative CT scans • Facial appearance and occlusal function.
The ischemia time and total operation time were shorter in the virtual planning group than in the conventional surgery group. High precision with the use of the cutting guides and templates was found for both the fibula and mandible, and a good fit was noted among the pre-bent plate, mandible, and fibula segments in the virtual planning group. Postoperative CT scans also showed excellent mandibular contours of the fibula flaps in accordance with virtual plans in the virtual planning group.
Surgical planning software (ProPlan CMF) was used preoperatively in the virtual planning group. In the virtual planning group, fibula flaps were harvested and osteotomized, and the mandibles were resected and reconstructed assisted by the prefabricated cutting guides and templates.
This study demonstrated that virtual surgical planning was able to achieve more accurate mandibular reconstruction than conventional surgery. The use of prefabricated cutting guides and plates makes fibula flap molding and placement easier, minimizes the operating time, and improves clinical outcomes.
Bias limitations: Medium risk of bias from the selection of participants, missing data, measurement of outcomes, and selection of reported results (retrospective review).

3D Virtual Planning
The aim/purpose of this study is to look at the impact of using 3D virtual surgical planning technology on surgical outcomes in the reconstruction of mandibular defects-specifically, type H mandibular defects including the condyle using a double-barrel vascularized fibula flap. The simulation allowed for the construction of an individual mandibular model serving to guide the clinical operation. They found preoperative virtual surgery greatly benefitted the actual surgery. In addition, postoperative 3D reconstruction revealed a close match with the simulated condyle. Therefore, combined virtual 3D reconstruction and rapid prototyping can improve postoperative outcomes in mandibular reconstruction.
Bias limitations: Medium risk of bias from the selection of participants, missing data, measurement of outcomes, and selection of reported results (retrospective review).

Limited Cone Beam CT (LCBCT)
The aim/purpose of the study is to look at the use of limited CBCT and subsequent 3D reconstruction to monitor bone resorption in alveolar bone grafts in patients with an alveolar cleft. They found that LCBCT scan and 3D reconstruction is a promising method for evaluation of the outcome of alveolar bone grafts.
One of the first and only uses of limited cone beam and grafting combination.
Bias limitations: Medium risk of bias from the measurement of interventions and departure from the intended interventions (retrospective review). 9. Heissler et al. Custom-made cast titanium implants produced with CAD/CAM for the reconstruction of cranium defects (17) Impact factor 3.87.

3D Spiral CT With CAD/CAM to Produce Custom-Made Titanium Implants
The aim/purpose of this paper was to use 3D spiral CT to design titanium implants for bony skull defects. They used rapid prototyping for a fine casting process-the first group to do this. One of the first papers to use spiral CT with CAD/CAM to create custom-made cast titanium implants for the reconstruction of cranium defects. They showed that this method is superior for several reasons for the patient and the surgeon: • Complex geometrical structures with small diameters can be produced with significantly more precision than previously.
• The rim of the implant can be designed in such a way that it extends over the edge of the bone -this leads to good stability at the bone-implant interface.
• The results are predictable and aesthetically very pleasing.

CAD Surgical Navigation
The aim/purpose of the paper was to see if the use of CAD and surgical navigation improved the surgical outcomes and operations in free fibula flap mandible reconstruction for patients with benign tumors who underwent primary unilateral reconstruction. They found that when including computer-aided design (CAD), CAD guided mandibular angle remodeling and condyle placement with increased accuracy. In addition, CAD and surgical navigation increase reconstruction accuracy without prolonging operative time.  (29) Impact factor 2.

MRI vs. CT
The aim/purpose of the study is to compare the accuracy of MRI based virtual 3D models of the intact orbit can approach that of the gold standard, CT-based models. This is to identify whether MRI is a viable alternative to CT scans in patients with isolated orbital fractures and penetrating eye injuries, pediatric patients, and patients requiring multiple scans in whom radiation exposure is ideally limited. Patients who presented with unilateral orbital fractures to a hospital in a 1-year period were recruited. The outcome measurements were orbital volume (primary outcome) and geometric intra-orbital surface deviations (secondary outcome) between the MRI-and CT-based 3D models.
The volumetric differences of the MRI models are comparable to reported results from CT models. The intra-orbital MRI surface deviations are smaller than the accepted tolerance for orbital surgical reconstructions. Therefore, the authors believe that MRI is an accurate radiation-free alternative to CT for the primary imaging and 3D reconstruction of the bony orbit.
Bias limitations: High risk of bias from the selection of participants, missing data, and measurement of outcomes (retrospective review).

Tenhagen et al. Three-Dimensional Handheld Scanning to
Quantify Head-Shape Changes in Spring-Assisted Surgery for Sagittal Craniosynostosis (20) Impact factor 2.8.

3d Handheld Scanning and Postprocessing Imaging Techniques
The aim/purpose of the study was to assess the utility of 3D handheld scanning photography in a patient group which underwent spring-assisted correction surgery for scaphocephaly. Usually, the gold standard for acquiring 3D imaging is CT that entails ionizing radiations and, in young children, a general anesthesia. 3D photographic imaging is an alternative method to assess patients who have undergone calvarial reconstructive surgery. They did Pre-and postop 3D scans acquired in theater, they then repeated these at the 3-week follow-up in clinic; images were then postprocessed for nine patients.
The following parameters were looked at: • Cephalic index (CI) • Head circumference • Volume • Sagittal length • Coronal width over the head Statistical shape modeling (SSM) was used to calculate the 3D mean anatomical head shape, no significant differences were observed in the CI between 3D and x-ray. Therefore, 3D handheld scanning followed by SSM proved to be an efficacious and practical method to evaluate 3D shape outcomes after springassisted cranioplasty in individual patients and the populationso a good alternative (or at least as a follow up) to CT.
Bias limitations: Medium risk of bias from the sequence generation, allocation concealment, and blinding (trial format study).

CONCLUSION
The field of facial imaging techniques for bony reconstruction can be seen to be clearly computed tomography heavy, but some small gains have been made in the magnetic resonance and handheld field.
One could think that with such a large analysis of the literature, it could serve as a guide for those looking to plan preoperatively. However, this would merely be sweeping generalizations of the technologies available. There are several things that are clear and should help when planning surgical intervention. The use of CT is the most commonly used modality for bony injuries and defects due to its availability, versatility (with modifications such as angiography) and accuracy. The emergence of technologies such as CAD lend themselves to using CT and provide an extra level of preoperative data. Similar to CT, MRI has many of these of these benefits, but for purely bone based planning it would not be the modality of choice. Depending on the situation of the patient (such as renal issues), MRI may end up being the image technique of choice and luckily it is robust enough to be of benefit. To make full use of the analysis carried out in this work, one should attempt to tailor their case to those papers with similar patient populations and intended outcomes.
The future of the technology is still unclear as it will rely on certain surgical technological milestones to be met (quicker image processing techniques and smaller devices) alongside user preferences. However, there is an ultimate patient benefit to being able to pre-plan procedures in both outcomes and time. There are still costs related to this and these techniques may be difficult to integrate into developing countries and centers without the equipment or user ability on site.

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.

AUTHOR CONTRIBUTIONS
ST: wrote majority and put paper together. MS: wrote parts of discussion and majority of introduction. RA: data extraction. ZJ and TD: data analysis and draft work. IW: oversaw project and draft work. All authors contributed to the article and approved the submitted version.