A Virtual Surgical Planning Algorithm for Delayed Maxillomandibular Reconstruction

Abstract

Background:

The absence of a tumor specimen, from which to obtain measurements at the time of delayed maxillo-mandible reconstruction, introduces degrees of uncertainty creating the need for substantial intraoperative guesswork by the surgeon. Using the virtual surgical planning(VSP) environment the size and shape of missing bony elements is determined, effectively “recreating the defect” in advance of the surgery. Three VSP techniques assist the reconstructive surgeon: patient-specific modeling, mirroring the normal contralateral side, and scaled normative data. To facilitate delayed reconstruction a hierarchical algorithm utilizing VSP techniques was developed.

Methods:

Delayed maxillo-mandible VSP reconstructions were identified from 2009–2016. Demographics, modeling techniques, and surgical characteristics were analyzed.

Results:

16 reconstructions were performed for osteoradionecrosis with displacement(50.0%) or oncologic defects(37.5%). Most patients had prior surgery(81.3%) and preoperative radiation(81.3%); 4 had failed prior reconstructions. The following delayed VSP techniques were used: patient-specific modeling based on previous imaging(43.8%), mirroring normal contralateral anatomy(37.5%), and scaled normative data(18.8%). Normative and mirrored reconstructions were designed to restore normal anatomy; however, most patient-specific VSP designs(71.4%) required non-anatomic reconstructions to accommodate soft tissue limitations and to avoid the need for a second flap. One partial flap loss required a second free flap, and one total flap failure occurred. Hardware exposure was the most common minor complication, followed by infection, dehiscence, and sinus tract formation.

Conclusion:

VSP has inherent advantages in delayed reconstruction when compared to traditional flap shaping techniques. An algorithmic approach based on available imaging and remaining native anatomy enables accurate reconstructive planning followed by flap transfer without the need for intraoperative guesswork.

INTRODUCTION

Over the last decade, there has been widespread expansion in use of virtual surgical planning (VSP) for reconstruction of a variety of osseous defects. The advantages of the VSP process over traditional techniques become pronounced when patients require delayed reconstruction of existent defects. Traditional techniques for shaping osseous flaps in head and neck reconstruction require assessment of the defect along with precise measurement of the tumor specimen.(1–3) In considering delayed reconstructions, normal anatomical landmarks are distorted from soft tissue scarring and/or radiation and the tumor specimen is no longer available. The need for precision remains, however. The delayed scenario requires substantially more guesswork for the surgeon when compared to immediate reconstructions; therefore, these defects are amenable to the potential advantages provided by the virtual surgical planning environment.(4–8)

A case series is presented of delayed reconstructions performed using a hierarchical algorithm of three computer-assisted modeling techniques. The study aim is to highlight to plastic surgeons how VSP technology can be used to eliminate the intraoperative guesswork associated with the traditional free-hand approach in this challenging subset of craniofacial defects.

METHODS

The study design is a retrospective review of delayed maxillary or mandibular osseous reconstructions performed with VSP between 2009–2016. Procedures were performed at two major academic medical centers. Demographic and surgical characteristics of the cohort were analyzed. VSP was performed as part of the standard of care at both institutions. 3D Systems (formerly Medical Modeling Inc.) Littleton, CO, assisted with the VSP process and cutting guide fabrication. Conceptual considerations and refinements in the VSP process were evaluated and determined by the senior authors (E.M. and E.G.). Internal Review Board approval was obtained from both institutions.

VSP reconstructive techniques were chosen at the discretion of the operating surgeon on a case-by-case basis according to the hierarchical algorithm shown in Figure 1. Defect location, size, anatomic distortion or constraints, and available imaging were taken into consideration. Cases of osteoradionecrosis were only included in the series if there was fracture of the mandible with significant displacement. Cases were further sub-classified as anatomic versus non-anatomic reconstructions depending upon whether the remaining anatomy would permit restoration to original proportions or if accommodations were needed for soft tissue scarring or radiation fibrosis. Once the dimensions of the defect were accurately re-established along with the optimal reconstructive plan, the operation was performed similar to any other immediate VSP reconstruction.

Figure 1.

Step-wise VSP technique algorithm for delayed reconstruction. The algorithm starts at the top and proceeds to lower tiers depending upon the data available and location of the defect.

Defining the pre-operative resection specifically for osteoradionecrosis involved a number of considerations distinct from other delayed reconstructive scenarios. Most important were the radiographically determined extent of disease and the level of disease/fistula formation on exam. First, we design the reconstruction to remove bone mesially and distally until the next anatomical mandible osteotomy is reached (parasymphysis, mid-body and angle). Next, two separate resections are planned, termed the narrow and wide margins, with two sets of mandible and fibula guides planned for manufacturing(6). Thereafter the VSP techniques described below are employed to guide the virtual reconstructive process. Finally at the surgery, the resection is performed until bleeding bone is visualized in accordance with the narrow and wide osteotomies. In none of the cases was there deviation from the preoperative VSP plans.

Patient-Specific

High quality imaging of the patient’s craniofacial skeleton prior to the trauma or resection serves as the anatomic reference for the virtual surgical plan. These cases have the greatest fidelity because the patients own morphometric data are used to guide the reconstructive process. In the absence of soft tissue constraints that preclude restoration of normal anatomy, this technique enables an accurate reconstructive design (Figure 2)

Figure 2.

Patient specific CT data was used to plan the delayed maxillary reconstruction. Patient had a limited soft tissue envelope with a contracted upper lip following radiation therapy (A). Barium (red) was painted onto the obturator prior to the CT scan in order to design the fibula reconstruction to fit the existing upper lip soft tissue envelope precisely. This would be considered a non-anatomic reconstruction (B). Fibula inset prior to upper lip closure (C). Postoperative result with the fibula (orange) overlaid with the surgical plan (blue) demonstrating precise execution (D).

Mirroring

For unilateral defects, a mirror image of the patient’s uninvolved contralateral side is inverted on screen in the virtual environment. It is then positioned over the defect, aligning it with remaining portions of the jaw or the temporomandibular joint. The osseous reconstruction is then virtually designed based on the mirrored normal anatomy. This technique is less reliable as defects become more anterior, especially if they extend across midline because no remaining uninvolved structure exists to mirror (Figure 3)

Figure 3.

The mirroring technique generates a mirror image reflection (yellow) upon the midline of the remaining normal anatomy. It is the repositioned at the defect, serving as a reference for the free fibula reconstruction (blue). Note the limitation of this technique in the setting of a large defect that extends anteriorly across midline.

Normative

While heterogeneity in mandibular size exists among individuals, the angles of the mandibular parabola at the parasymphysis, mid-body, and angle, are relatively constant or preserved. (7) Therefore, a standard anatomic template can be utilized for reference after it has been adjusted by scaling up or down in size in the virtual environment to match the patient’s mandible (Figure 4). An example of delayed lateral mandible reconstruction using normative data and a scapula flap is demonstrated in Figure 5 and 6. The normative method can also be particularly useful for anterior defects (Figure 7). Importantly, because normative data uses none of the patient’s own anatomy it represents the lowest tier on the delayed VSP hierarchical algorithm. As such, cases planned with this technique in particular must be critically evaluated to ensure that the design adequately restores facial shape and normal occlusion.

Figure 4.

Normative data takes advantage of preserved angles between mandibular anatomic segments to generate a virtual reference for reconstruction (green), however, it must be adjusted to the appropriate scale based on patient size (red).

Figure 5.

Collapse of right lateral mandible following fracture of a reconstruction plate (left image). The size of the original segmental defect can be appreciated through overlay of a normative green mandible (center). VSP proceeds like any other immediate reconstruction based on the corrected defect size (right).

Figure 6.

Remaining portions of the right mandible body are planned to be excised to healthy thicker bone stock using VSP and intraoperative cutting guides (top). Preoperative and intraoperative photos (bottom left images) demonstrate the fractured plate and bony defect with displacement of the distal and mesial mandible segments. Collapse of the segments on each other makes determination of the original defect unreliable. Note the segments orientation to each other is reversed because of the mobility of the ramus. Intraoperative and long-term postoperative images of the reconstructed mandible (bottom right images).

Figure 7.

Comparison of mirroring (yellow) and normative (red) VSP techniques for defects that extend anteriorly across midline.

RESULTS

Over the seven-year study period, 16 complex maxillary and/or mandibular osseous reconstructions with vascularized bone were performed in a delayed fashion in 15 patients using virtual surgical planning. Patient demographics are listed in Table 1.

Table 1.

Patient Demographics

Patients	n=16
Age	51.7 ± 14.7
Gender	9 male, 7 female
Prior Surgery	13 (81.3%)
Prior Reconstruction	4 (25.0%)
Preoperative XRT	13 (81.3%)
Follow up	20.9 ± 17.7 months
Diagnosis
ORN	8 (50.0%)
Oncologic	6 (37.5%)
GSW	2 (12.5%)

There was a male predominance (60%) with a mean age of 51.7 years (range 28–76). Most reconstructions were performed for either osteoradionecrosis (ORN) with displacement/distortion (50.0%) or oncologic defects (37.5%); two patients sustained gunshot wounds to the face (12.5%). The majority of patients had undergone prior surgery (81.3%) and preoperative radiation (81.3%). Four (25%) patients had failed prior reconstructions. Mandibular defects were most common (81.2%), followed by the maxilla (12.5%), and one combined maxillary and mandibular defect (6.3%). Free fibula flaps were used in all but one case for vascularized bone, with an average of 2.38±0.72 segments per flap. Flaps for anterior mandible defects had a greater number of segments 2.83±0.41 compared to lateral mandible defects 1.71±0.49 (p<0.01). All three maxilla reconstructions required 3 bone segments.

Patient-specific modeling was used most commonly (n=7, 43.8%), followed by mirroring normal contralateral anatomy (n=5, 31.3%) and finally scaled normative data (n=4, 25.0%). The VSP techniques used according to each defect are listed in Table 2. Normative and mirrored reconstructions were designed to restore normal anatomy. However, all but two of the patient-specific reconstructions (71.4%) necessitated non-anatomic reconstructions related to limitations imposed by adjacent native bone and/or soft tissue. No virtual surgical plans were aborted intra-operatively. The mean follow up length was 20.9 months (range 10–38).

Table 2.

Virtual Surgical Planning Technique According to Reconstruction

VSP Method	Diagnosis	Location	Defect Subtype	Anatomic	Flap	Recipient Artery	Recipient Laterality	Bone Segments
Normative	Osteosarcoma	Mandible	Anterior, Lateral	Y	Fibula	Facial	Ipsilateral	3
Normative	SCC	Mandible	Anterior, Subtotal	Y	Fibula	External Carotid	Right (n/a)	3
Normative	ORN	Mandible	Lateral	Y	Fibula	Facial	Contralateral	2
Normative	SCC	Mandible	Lateral	Y	Scapula	Facial	Ipsilateral	1
Mirroring	ORN	Mandible	Lateral	Y	Fibula	Superior Thyroid	Ipsilateral	2
Mirroring	SCC	Maxilla	Type III	Y	Fibula	Superficial Temporal	Ipsilateral	3
Mirroring	GSW	Mandible/Maxilla	Lateral, II	Y	Fibula	Facial	Ipsilateral	3
Mirroring	GSW	Mandible	Lateral	Y	Fibula	Superior Thyroid	Ipsilateral	2
Mirroring	ORN	Mandible	Lateral	Y	Fibula	Transverse Cervical	Ipsilateral	1
Patient Specific	SCC	Maxilla	Bilateral Type III	N	Fibula	Lingual	Left (n/a)	3
Patient Specific	ORN	Mandible	Lateral	N	Fibula	Facial	Ipsilateral	2
Patient Specific	Ameloblastic Carcinoma	Mandible	Anterior, Lateral	N	Fibula	Facial	Contralateral	3
Patient Specific	ORN	Mandible	Anterior, Subtotal	N	Fibula	Facial	Right (n/a)	3
Patient Specific	ORN	Mandible	Anterior, Subtotal	N	Fibula	Lingual	Left (n/a)	3
Patient Specific	ORN	Mandible	Lateral	Y	Fibula	Facial	Contralateral	2
Patient Specific	ORN	Mandible	Anterior, Subtotal	Y	Fibula	Lingual	Left (n/a)	2

Maxillectomy: Type II ‐ hemimaxillectomy with orbital floor preserved; Type III – hemimaxillectomy with preservation of orbital contents

Reconstructive outcomes are listed in Table 3. Four flaps (25%) suffered from perioperative vascular compromise requiring return to the operating room for exploration, with one partial flap loss requiring a second free flap (6.3%), and one total flap failure (6.3%). Hardware exposure was the most common minor complication (3), followed by infection (2), dehiscence (2), and sinus tract formation (2). Five flaps underwent revisionary procedures and two patients achieved dental restoration with osseointegrated implants.

Table 3.

Reconstructive Outcomes

Complications
Major
Takeback	4
Partial flap failure	1
Total flap failure	1
Minor
Hardware exposure	3
Infection	2
Wound dehiscence	2
Chronic sinus tract	2
Oncologic Recurrence	2
Flap Revision	5
Dental Restoration	2

DISCUSSION

In order to effectively restore normal and functional anatomy, a top priority should be a precise and accurate reconstruction. While traditional methods of vascularized bone flap design are capable of reliably addressing the majority of craniofacial defects, when anatomic reference points are missing or distorted such as in delayed reconstruction, the short-comings of the free hand technique become more obvious. The key to success in delayed reconstruction is recreation of the original defect; however, using the free hand technique there is significant trial and error because the surgeon cannot reliably estimate the initial defect in absence of a specimen and in the presence of soft tissue contracture. In contrast, the virtual surgical planning(VSP) environment facilitates precise determination of the size and shape of missing bony elements, more effectively “recreating the defect,” and doing so in advance of the surgery. Intraoperatively, the surgery proceeds akin to an immediate rather than a delayed reconstruction because of the assistance of customized cutting guides that obviate the need for traditional landmarks and enable execution of the virtual plan..(9–12) Importantly, the actual osseous construct becomes an added “known” reference point because of its reliability in size and shape when created using VSP. Although gross visual inspection of the final reconstruction may appear no different when compared to the traditional method of bone shaping, the VSP reconstruction is performed on an anatomically precise defect more closely resembling the premorbid state.

There is an assumption that application of VSP in reconstruction of osseous defects is straightforward. However, implementation and adoption of new technology is associated with a learning curve that can be more rapidly overcome through knowledge conveyed by colleagues. Thus, one premise of the current report is to disseminate information about the application of VSP for delayed repair of maxillomandibular defects using an algorithm developed at two academic medical centers with significant experience. The algorithm described is distinguished from classification schemes because it enables an unambiguous conceptual approach to address any delayed defect. As one proceeds from top to bottom, less of the patients own anatomy is used to model the reconstruction, and therefore it is deemed less accurate. Although some of the individual techniques described may already be in use by some, no literature exists describing a systematic approach using CAD technology to reliably reconstruct the current set of defects. Moreover, surgeons who are new to the VSP process can use the described algorithm to allow for a systematic approach to some of the most complicated reconstructions.

The patient-specific technique is primarily used in oncologic or ORN cases, since it is unusual for trauma patients to have pre-injury imaging studies available. This approach allows for design of the optimal reconstruction taking into account the anatomic limitations of each individual defect. These cases are further subdivided into anatomic versus non-anatomic reconstructions depending upon whether or not the patient’s anatomy was completely restored. As illustrated in Table 2, the majority of cases using this approach actually required non-anatomic adjustments.

Looking specifically at the five non-anatomic reconstructions using the patient-specific technique, four of them were designed to result in an under-projected anterior mandible or maxilla due to restrictions of the soft tissue envelope from scarring or radiation. Although the original anatomy is adjusted, it still uses the patient’s own anatomical proportions especially when compared to the normative approach. This method also avoids the need for a second free flap for external soft tissue coverage (13) along with the identification of an additional set of recipient vessels, as well as optimizes aesthetic outcomes by eliminating an external patch-like appearance. The fifth patient had severe osteoradionecrosis of the mandibular body bilaterally with a unilateral pathologic fracture. This reconstruction was designed with reduced mandible proportions as it was felt that attempting to restore an anatomically correct position would potentially result in fracture of the contralateral diseased mandible.

The second tier of the algorithm also uses the patient’s own anatomy. Mirroring is a simple and reliable method to re-establish normal anatomy for unilateral defects. On-screen manipulation of the uninvolved side in the virtual environment can, for example, allow correct repositioning of a lateral ramus segment that has been pulled towards the midline because of unopposed action by the pterygoids. However, for anterior defects, especially those extending across the midline, no mirrored structure exists so this technique becomes less precise with an element of approximation required.

The lowest tier of the delayed VSP reconstruction algorithm uses normative data stored by the surgical planning vendor. Prior studies have demonstrated that the shape, but not the size, of the mandible is highly conserved amongst individuals. (7) The traditional osteotomy angles, performed for mandible reconstruction, vary by less than 5 degrees among individuals at the parasymphysis, midbody, and angle. Normative data, adjusted to scale takes advantage of this phenomenon.

Another area where VSP appears to have substantial advantages over traditional shaping techniques is in maxillary reconstruction. The compact and highly three-dimensional architecture of the maxilla results in a significant reconstructive challenge. Multiple acute osteotomies of small bone segments are required as the shape of the maxilla changes significantly within a limited amount of space, especially when compared to the relatively gradual changes seen in the mandible.

The current study has some limitations, including its retrospective nature. The small cohort size reflects the low incidence of defects with the degree of reconstructive difficulty required for inclusion in this series. Furthermore, the unique and complex nature of each case did not enable a meaningful comparison with a matched control group reconstructed using traditional shaping techniques. Therefore, demonstration of superiority of traditional versus VSP techniques was not intended, nor possible. Lastly, the inherently complex nature of delayed reconstructions, including factors such as previous surgery, vessel-depleted necks, and prior radiation, may partially explain the high rate of vascular compromise observed. Although there were 4 reoperations for vascular compromise, three were salvaged with only 1 complete flap loss.

In conclusion, experience with this complicated subset of delayed maxillary and mandibular pathology highlights technical advantages of VSP. The ability to accurately design an optimal reconstruction and precisely execute it in the absence of normal anatomic landmarks constitutes a powerful modern surgical tool not previously available in the reconstructive armamentarium.