Are Skin Fiducials Comparable to Bone Fiducials for Registration When Planning Navigation-assisted Musculoskeletal Tumor Resections in a Cadaveric Simulated Tumor Model?

Abstract

Background

To improve and achieve adequate bony surgical margins, surgeons may consider computer-aided navigation a promising intraoperative tool, currently applied to a relatively few number of patients in whom freehand resections might be challenging. Placing fiducials (markers) in the bone, identifying specific anatomical landmarks, and registering patients for navigated resections are time consuming. To reduce the time both preoperatively and intraoperatively, skin fiducials may offer an efficient and alternative method of navigation registration.

Questions/purposes

(1) Does preoperative navigation using skin fiducials for registration allow the surgeon to achieve margins similar to those from bone fiducial registration in a simulated lower extremity tumor resection model in cadavers? (2) Does the use of preoperative navigation using skin fiducials for registration allow the surgeon to achieve similar bony margins in pelvic resections of simulated tumors as those achieved in long-bone resections using only skin fiducials for navigation in a cadaver model?

Methods

Simulated bone tumor resections were performed in three fresh-frozen cadavers with intact pelvic and lower-extremity anatomy using navigation guidance. We placed 5-cm intraosseous cement simulated bone tumors in the proximal/distal femur (n = 12), and proximal/distal tibia (n = 12) and pelvis (supraacetabular; n = 6). After bone tumor implantation, CT images of the pelvis and lower extremities were obtained. Each planned osseous resection margin was set at 10 mm. Navigation registration was performed for each simulated tumor using bone and skin markers that act as a point of reference (fiducials). The simulated bone tumor was resected based on a resection line that was established with navigation, and the corresponding osseous margins were calculated after resection. These margins were determined by an orthopaedic surgeon who was blinded to resection planning by the removal of cancellous bone around the cement simulated tumor. The shortest distance was measured from the cement to the resection line. Smaller mean differences between planned and postoperative margins were considered accurate. Independent t-tests were conducted to assess measurement differences between planned and postoperative margins at the 95% CI. Bland-Altman analyses were conducted to compare the deviation in margin difference between planned and postoperative margins in skin and bone fiducial registration, respectively.

Results

In all, 84 total resection margins were measured with 48 long bone and 20 pelvic obtained with skin fiducials and 16 long bone obtained with bone fiducials. The planned mean margin was 10 mm for all long bone and pelvic resections. We found that skin fiducial and bone fiducial postoperative margins had comparable accuracy when resecting long bones (10 ± 2 mm versus 9 ± 2 mm, mean difference 1 [95% CI 0 to 2]; p = 0.16). Additionally, skin fiducial long bone postoperative margins were comparable in accuracy to pelvic supraacetabular postoperative margins obtained with skin fiducials (10 ± 2 mm versus 11 ± 3 mm, mean difference -1 mm [95% CI -3 to 1]; p = 0.22). When comparing the deviation in margin difference between planned and postoperative margins in skin and bone fiducial registration, 90% (61 of 68) of skin fiducial and 100% (16 of 16) bone fiducial postoperative margins fell within 2 SDs.

Conclusions

In this pilot study, skin fiducial markers were easy to identify on the skin surface of the cadaver model and on CT images used to plan margins. This technique appears to be an accurate way to plan margins in this model, but it needs to be tested thoroughly in patients to determine if it may be a better clinical approach than with bone fiducials.

Clinical Relevance

The margins obtained using skin fiducials and bone fiducials for registration were similar and comparable in this pilot study with a very small effect size. Boundaries of the simulated tumors were not violated in any resections. Skin fiducials are easier to identify than bone fiducials (anatomic landmarks). If future clinical studies demonstrate that margins obtained using skin fiducials for registration are similar to margins obtained with anatomical landmarks, the use of navigation with skin fiducials instead of bone fiducials may be advantageous. This technique may decrease the surgeon’s time used to plan for and localize registration points and offer an alternative registration technique, providing the surgeon with other registration approaches.

Introduction

Background

Primary malignant tumors, metastatic tumors, and aggressive benign tumors located in the long bones and pelvis are best managed with tumor resection with adequate margins and documentation of these margins. Adequate margins have been shown to be associated in some patients undergoing resection for tumor with a decrease in the likelihood of local recurrence and improved long-term survival [3, 15, 20]. Bone tumor resection can be challenging even for experienced surgeons because of the complex anatomic locations of the tumor, especially in the pelvis. In particular, the interpretation of preoperative images and the lack of intraoperative guidance to confirm and document tumor resections is difficult. Relying solely on radiography or fluoroscopy to identify bone tumor margins intraoperatively is often not adequate [1, 18, 19]. In the pelvis, complex 3-D anatomy makes it difficult to resect tumors and obtain negative margins. In the setting of image interpretation and complex neurovascular structures, the ability to achieve negative surgical margins often depends on the surgeon’s experience and frequently involves more extensive surgical exposure [1, 19]. To achieve more accurate surgical bony margins in patients undergoing complex resections, surgeons consider computer-aided navigation a promising intraoperative tool. With computer-aided resections, surgery can be planned preoperatively and intraoperatively [10, 24].

Navigation systems are currently used for pedicle screw applications, cup placements for hip arthroplasties, cruciate ligament reconstructions, and fixation of extremity and pelvic fractures [1, 2, 7, 10, 13, 24]. Recent advancements in various aspects of orthopaedic surgery have opened a new era in orthopaedic surgical planning and management of musculoskeletal neoplasms, especially in the pelvis. Current navigation systems use CT, MRI, or a combination of both imaging modalities. The fusion of MRI and CT images for intraoperative navigation may increase the accuracy of these resections [12, 14, 21, 22, 24]. Accurate navigation registration is the cornerstone of performing intraoperative navigation and resection accurately and precisely.

Rationale

Navigation registration is defined as the process of synchronizing preoperative imaging with the anatomic landmarks on the patient undergoing surgery, which uses anatomic point recognition to match these points with their images (image-to-patient registration; MRI or CT). Selecting and locating anatomic points intraoperatively for navigation registration may increase operative time [9, 25], and may also increase the radiation exposure if exact points cannot be identified because of variables involving the operating room environment [11]. It may lead to potential errors of intraoperative mapping and thus the incorrect orientation of the patient’s surgical area [6, 16, 23] when anatomic landmarks are not perfectly (exactly) located. Skin fiducials are skin markers commonly used in diagnostic radiology and may be easier and quicker to apply preoperatively and to anatomically locate intraoperatively, potentially reducing radiation exposure. Registration challenges with anatomic landmarks (bone fiducials) motivated us to test whether a less time-consuming registration process and similar accuracy as bone anatomic landmark registration could be performed using skin fiducials, or markers that act as a point of reference that are easy and quick to identify, in CT or MRI imaging preoperatively and intraoperatively.

Study Questions

(1) Does preoperative navigation using skin fiducials for registration allow the surgeon to achieve margins similar to those from bone fiducial registration in a simulated lower extremity tumor resection model in cadavers? (2) Does the use of preoperative navigation using skin fiducials for registration allow the surgeon to achieve similar bony margins in pelvic resections of simulated tumors as those achieved in long bone resections using only skin fiducials for navigation in a cadaver model?

Materials and Methods

Study Design and Setting

This cadaver pilot study was ruled exempt from institutional review board approval by the University of Washington Medical Center’s institutional research board. The navigation system we used for this study was the Navigation System II with OrthoMap 3D navigation software (Stryker, Kalamazoo, MI, USA).

Specimens

Simulated bone tumors were resected in three fresh-frozen cadavers with intact pelvic and lower-extremity anatomy using navigation guidance. None of the cadavers used in this study had prior pelvic or lower extremity surgical procedures.

Description of Experiment, Treatment, or Surgery

Skin and bone fiducials for registration were placed in similar extremity and pelvic locations (Fig. 1). Skin fiducial markers (Multi-Modality Radiology Markers, PDC Healthcare, Valencia, CA, USA) with adhesive backing were placed on the skin surface of the tibia, femur, and pelvis. We placed 5-cm intraosseous simulated bone tumors (cement mass) in the proximal and distal femur (n = 12), and proximal and distal tibia (n = 12), and pelvis (supraacetabular; n = 6). An open approach was used where an incision was made and a defect in the cortex was created. A roughly 5-cm cement tumor was created, measured, and placed within the cortical defect. After the tumor was placed, the incision was closed. Incision sites were chosen such that they would not interfere with conventional surgical approaches for tumor resection.

Fig. 1.

This picture shows the position of the skin fiducials in the lower extremities and pelvis. The position was repeated for every specimen.

Bone fiducials were created by using anatomic landmarks and then placing 2.0-mm K-wires in bone. They were positioned proximal and distal to the simulated tumor. To ensure accurate navigational system registration, we placed four fiducials for each segment to resect. CT images of the pelvis and lower extremities were obtained following the Stryker recommended protocol for navigation using 1-mm cuts. These CT images were loaded into the navigation system to plan preoperative margins after tumors were implanted. Each planned osseous resection margin was set at 10 mm for all proximal and distal long-bone margins and all superior, inferior, anterior, and posterior, osseous margins in the pelvis (Fig. 2). Soft tissue margins were not included. When planning and resecting, if a 10 mm margin was not possible given the simulated tumor’s proximity to the end of a bone or cortex or a margin involving soft tissues, the margin was considered missing and was not included in the analysis. At the time of tumor resection, skin or bone fiducials were used for registration before surgical approaches were made. Both skin and bone fiducials were identified in the cadaver through a simple inspection and localization on preoperative CT images (Fig. 3).

Fig. 2 A-D.

Axial and coronal CT views of preoperative planned resection planes with a 5-cm cement simulated tumor. (A) Axial view of two planned margins (10 mm) for a cement tumor located in the pelvic supra-acetabular region. (B) Coronal view of two planned margins (10 mm) for the same supra-acetabular cement tumor. (C). Two margins (10 mm) were planned for this proximal femur cement tumor (D) Two margins (10 mm) were planned for this distal femur cement tumor.

Fig. 3.

3-D reconstruction of the CT focused on the cadaveric right lower extremity shows the skin fiducials and their location on the thigh (indicated by labels 1-6). The location of the fiducials was consistent across cadaveric specimens and are easily identifiable on CT imaging utilized in planning the resection. Proximal and distal transverse lines indicate the line of resection that was planned in bone for proximal femur resection and distal femur resection.

The navigation system was stationed 4 feet to 5 feet from the specimen with its infrared receiver angled down at 45°. Patient trackers were rigidly fixed to the bone for each resection which communicated with the navigation’s infrared receiver. In all specimens, trackers were fixed at least 5 cm away from skin fiducials that would be used to avoid skin deformation. Navigation system registration was performed for each simulated tumor using the bone or skin fiducial. Registration was completed through a process called image-to-patient registration, where preoperatively determined surface points (4 points to 5 points) from fused images are located on preoperative imaging and the corresponding surface point is located on the patient [24]. The Stryker OrthoMap 3D protocol required that the navigation system registration possess a mean error of less than 3.0 mm. Using a navigation infrared pointer, the identification and registration of skin fiducials was performed before any incision was made on the skin. The navigation pointer was placed in the center of each skin fiducial. We obtained bone fiducial registration by placing the navigation pointer superior to where the K-wire and bone cortex intersected. After computer-based navigation system registration was completed without using bone surface refinement technique (adding additional points of interest to the registration field), and registration accuracy was confirmed to be < 1 mm, simulated bone tumor resection was initiated. Tumors were resected using classical anterior surgical approaches in accordance with the anatomic segment. A navigation infrared pointer was used to identify the planned surgical margins in the bone to resect. A planned resection line was identified and marked on the bone. The bone was resected based on the resection line that was established with navigation. Every resection was carried out using a 1.37-mm oscillating saw. Additional intraoperative imaging was not used because the purpose of this pilot study was to assess whether skin fiducials may be an alternative to anatomic landmark registration.

Methods used in the present study were chosen to simulate a more realistic setting in which simulated patients have a bone tumor deemed appropriate for resection. Anatomic landmarks or skin fiducials were identified in the images preoperatively as they are currently identified as part of the patient’s clinical course. The navigation pointer marked resections in bones and guided the direction of resections.

Variables, Outcome Measures, Data Sources, and Bias

Planned margin data points were retrieved from the computer-based navigation system and postoperative margins were obtained via manual measurement by the surgeon (RZ). Computer-based margins were planned, and resection approaches were routine: anterior in the extremity and ilioguinal in the pelvis. The cadaver-based resections were performed between March 2016 and June 2016 in a laboratory setting. The postoperative margin was defined as the shortest distance from the osteotomized cortex to the tumor cement’s edge. These margins were determined by a different orthopaedic surgeon (CY) who was blinded to the resection planning. Margins were defined as the shorter distance from simulated cement tumor to the resection line. To measure the margin, the simulated cement tumor was exposed, cancellous bone was removed, and the margin distance was measured with a depth caliper and recorded. The surgeon (RZ) determined accuracy by comparing the postoperative margins with planned resection margins obtained from the computer-based navigation system.

Statistical Analysis, Study Size

We used the Statistical Package for Social Sciences version 24.0 (IBM Corp, Armonk, NY, USA). Independent t-tests were conducted to assess the measurement differences between the planned and postoperative margins at the 95% CI. We conducted Bland-Altman analyses to compare the deviation in margin difference between the planned and postoperative margins in skin and bone fiducial registration, respectively. When most deviations in margin differences are within 2 SDs of the mean value, a technique is classified as having good agreement [4]. When planned and postoperative margins are described, smaller mean differences between these values were termed as accurate. Values presented were rounded to the nearest millimeter because submillimeter values are difficult to clinically conceptualize.

Accounting for all Specimens

Sixteen long-bone simulated tumor resections using skin fiducials and eight long-bone simulated tumor resections using bone fiducials were performed. Anatomic landmarks (bone fiducials) are considered to be standard for registration [25], which is why fewer simulated tumor resections using bone fiducials were performed. Six pelvic supraacetabular simulated tumor resections were performed using skin fiducials. Pelvic resections were not included in the bone fiducial group due to specimen availability. Eighty-four total resection margins were obtained and measured with 48 long bone and 20 pelvic supraacetabular margins obtained with skin fiducials and 16 long bone margins obtained with bone fiducials. Planned mean margin was 10 mm for all long bone and pelvic resections. When a 10-mm margin was out of bone, the margin was considered missing and not included in the analysis. The mean postoperative margins for long-bone resections conducted with skin fiducials was 10 ± 2 mm and with bone fiducials was 9 ± 2 mm. The mean postoperative margin for pelvic supraacetabular resections conducted with skin fiducials was 11 ± 3 mm (Table 1). All simulated resections had registration errors of < 1mm.

Table 1.

Mean postoperative margins obtained with skin and bone fiducials by resection location

Pelvic resections using skin fiducials (n = 20), with mean postoperative margin 11 ± 2 mm
Specimen 1
Resection location	Pelvis
Margin location	Superior	Inferior	Anterior	Posterior
Measured postoperative margin (mm)
Right side	10	9	10	13
Left side		11
Resection location	Distal tibia		Proximal tibia		Distal femur		Proximal femur
Specimen 2
Resection location	Pelvis
Margin location	Superior	Inferior	Anterior	Posterior
Measured postoperative margin (mm)
Right side	10	15	14
Left side	15	15	16	5
Specimen 3
Resection location	Pelvis
Margin location	Superior	Inferior	Anterior	Posterior
Measured postoperative margin (mm)
Right side	8	7	17	9
Left side	10	12	12	8

Other Methods

We did not use intraoperative imaging in our study. The navigation system was the only guidance for the boney resections. This pilot study lacked a control group that did not use navigation assistance for osseous resections.

Results

Skin Versus Bone Fiducials

We found that skin fiducial and bone fiducial postoperative margins had comparable accuracy when resecting long bones (10 ± 2.0 mm versus 9 ± 2 mm, mean difference 1 [95% CI 0 to 2.0]; p = 0.16). When comparing the deviation in margin difference between planned and postoperative margins in skin and bone fiducial registration, 90% (61 of 68) of skin fiducial and 100% (16 of 16) bone fiducial postoperative margins fell within 2 SDs, indicating that both registration types had good agreement.

Skin Fiducial Long-bone versus Pelvic Margins

Skin fiducial long bone postoperative margins had accuracy comparable to skin fiducial pelvic resections (10 ± 2 mm versus 11 ± 3 mm, mean difference -1 mm [95% CI -3 to 1 mm]; p = 0.22).

Discussion

Background and Rationale

A challenge of computer-based navigation in complex surgical resection is the ability of the software registration to consistently and accurately identify anatomic structures on preoperative imaging and to reproduce these images in the operating room environment [6, 11, 16, 23]. Additionally, our previous experience with challenging pediatric tumor resections led our team to create this pilot study as a way to test an alternative approach to the navigation registration process. The skin fiducials appeared to be accurate in planning margins that were comparable to bone fiducials in this cadaver model of simulated bone tumors. Within the skin fiducial technique, long bone and pelvic supra-acetabular margins were also found to be comparable.

Limitations

Although we were able to assess the functionality and potential impact that skin fiducials could produce in computer-based surgical navigation, the study has many limitations. First, this pilot study used simulated tumors in a cadaver model. It is possible that skin fiducials as we tested them will not work the same in patients who are obese, who have differing skin thicknesses, and in patients who must be repositioned intraoperatively. Different types of skin fiducials or alternative minimally invasive registration techniques may be needed to achieve surgical registration in patients of differing BMIs and patients who may be repositioned during the operation. In our cadaver model, we positioned our specimens on the operative table in the same position that the specimen was placed when the CT imaging was performed. Additionally, for this modality of registration it is essential that the incisions are made after the skin fiducial registration step was accomplished. The surface refinement technique may help improve registration accuracy, but we were not sure how the Stryker algorithm would respond to skin-based surface refinement, so we left it out of our testing model. Once the tracker is fixed in bone and registration is finalized, the extremity and skin can be mobilized freely with no risk of changing navigation parameters. In our simulated resections we did not re-register.

Future studies will need to address whether re-registration may be viable in this registration type or whether this presents another limitation of skin fiducial registration.

Secondly, in this study we used registration techniques that do not reflect clinical use and this pilot study may not be directly comparable with the live patient situation. Clinically, intraoperative imaging can be done to automatically register the patient to imaging within the operating room; however, this can lead to additional radiation exposure, which may be more of a concern in pediatric patients, whom we had in mind when designing this study. If skin fiducial registration is to be considered an alternative to anatomic landmark or intraoperative imaging registration, sterilization of skin fiducials must be addressed.

Our study purpose was to test a new registration technique that may help direct other investigators testing alternative ways to approach complex tumor resections. In this study, we did not directly compare pelvic supraacetabular resections using skin fiducials versus bone fiducials for two reasons: We had limited cadaver resources, and we assumed that anatomic landmarks are the standard for navigation. Skin fiducials for registration in the pelvis had minimal variance and were able to produce postoperative margins that fell within the Stryker acceptable 3-mm threshold. Additionally, this study was unable to control for mean measurement differences due explicitly to human error. Future studies must compare the skin and bone fiducial methods in the pelvis, address how to standardize registration techniques, and compare these fiducial types in other models before these results can be directly translated into clinical practice. Second, the simulation of bone tumors with cement simplified the assessment of postoperative margins and measurements. Malignant osseous tumor resections are rarely as well defined as the current bone tumor model used in this study. Future studies must assess the impact of other registration techniques in live patients to determine which technique may help surgeons achieve adequate margins in patients undergoing complex surgical resections and provide the patient with negative surgical margins.

Skin Versus Bone Fiducials

This study showed that skin fiducial registration was a method that allowed planning and achieving accurate postoperative margins in a simulated tumor model in cadavers that was comparable to registration using bone fiducials for long-bone resections. Although the effect size is small in this pilot study, both methods led to the ability of the planned resections to result in negative margins. Additionally, we observed good measurement agreement: 90% (61 of 68) of skin fiducial and 100% (16 of 16) bone registration margin difference deviations were within 2 SDs of the mean value. This study had a small effect size and the differences in skin and bone fiducial registration were very small. This finding is still important because skin fiducials may provide an equivalent registration method for bone tumor resections. Clinically, anatomical landmarks and or intraoperative imaging are used, however, skin fiducial registration has the potential benefit of being easier to identify during preoperative planning, in the registration process, and intraoperatively. With skin fiducials there is no need to identify anatomic landmarks before surgery, which can be difficult to localize intraoperatively, and they may provide less radiation exposure to a patient, thus reducing barriers to using surgical navigation. In a previous navigation study examining the mean difference between planned and postoperative margins in a clinical sample of individuals with osseous sarcoma, Stoll et al. [23] reported that the mean difference exceeded the 3.0-mm limit. In a static cadaveric model, Eccles et al. [8] evaluated the measurement agreement between planned and resected surgical margins of simulated soft-tissue tumors as a follow-up study. The authors reported a mean margin difference of 0.75 mm, within the 3.0 mm acceptable error of the navigation system. Although both the skin and bone fiducial techniques of the present study provide mean differences between the planned and postoperative margin that is within acceptable limits, this experiment was conducted in a controlled research environment using cadavers, meaning that direct translation into a clinical setting should be confirmed by additional investigation.

Skin Fiducial Long-bone Versus Pelvic Margins

Mean postoperative margins obtained with skin fiducial registration in long-bone resections were as accurate as pelvic supraacetabular resections. Mean postoperative margins obtained in this pilot study were consistent with previously published navigation studies reporting small differences between planned and obtained surgical margins [8, 9, 25]. In this study, we expected a larger difference between planned and mean postoperative margins in the pelvis compared with long bones as the pelvis has complex anatomy, multiplanar resections, and can involve challenging surgical approaches. Further, pelvic resection and reconstruction remains challenging even with ideal surgical conditions [5]. Although we found comparable mean postoperative margin differences between long-bone and pelvic supraacetabular resections in a cadaveric model, our effect size was small and future studies must determine whether skin fiducial registration may provide an alternative for patients who are deemed appropriate for navigation technique in challenging clinical resections.

The margins obtained in our study are comparable with the margins obtained in previously published reports on tumor resection of the long bones and pelvis [21, 22, 24]. An increase in margin deviation in the pelvis is expected because of more complex anatomy, multiplanar resections, and challenging surgical approach, however, when comparing skin and bone registration techniques, we reported comparable margin differences. Cartiaux et al. [5] noted that resection and reconstruction of the pelvis remained challenging even within ideal surgical conditions. Specifically, surgeons had a 52% probability of achieving a 10.0-mm free-hand pelvic surgical margin within ± 5 mm tolerance. Alternatives have been described to improve the accuracy of resections through navigation. Intraoperative imaging or even patient-specific instruments may improve resection accuracy while decreasing the difference between planned and resected margins [12, 14, 17, 18, 25]. However, the use of intraoperative imaging increases the patient’s exposure to radiation and increases the costs of these procedures [17]. Although the present pilot study has a small effect size, skin fiducial registration may provide a navigation registration technique that may decrease variability in surgical margins with the benefits of limiting radiation exposure and ease of use. How skin fiducial registration will work in a clinical scenario awaits future investigation.

Conclusions

In this pilot study using a cadaver model, skin fiducial markers were easy to identify on the skin surface and on CT images used to plan margins. While effect sizes are small in this study, mean postoperative margins obtained with skin fiducial registrations of both long bone and pelvic supraacetabular resections were comparable. We also found that mean postoperative margins obtained while using skin fiducial registration in long-bone and pelvic supraacetabular resections were comparable. There was no violation of boundaries of the simulated tumors in any resections. Skin fiducial markers are available for both CT and MRI. They are easy to identify on images and remain in place during registration. This technique appears to be an adequate way to plan margins in this cadaveric model, but it must be tested thoroughly in patients to determine if it may be a better clinical approach than with bone fiducials. If future clinical studies demonstrate that margins obtained using skin fiducials for registration are similar to margins obtained with anatomical landmarks (bone fiducials), the use of navigation with skin fiducials may be advantageous because it may help facilitate complex surgical resections.