Accuracy and limitations of computer-guided curettage of benign bone tumors

Abstract

Adequate curettage of benign bone tumors located close to articular joints or neurovascular tissue is difficult without damaging those tissues. The purpose of this study was to evaluate the adequacy of tumor removal in computer-assisted curettage of benign bone tumors. The study is a prospective case series involving eight patients with benign bone tumors located near an articular joint or major neurovascular tissue. Image-to-patient registration with the navigation system was performed using paired-points methods in conjunction with CT images. A cortical window was created to visualize the tumor cavity. After removal of the gross tumor with sharp curettes, a specially designed burr attached to a navigation probe was used to monitor the location of the burr tip in real time. The high-speed burr extended the bony margin a few millimeters over the cavity wall. The empty cavity was then filled with bone cement. We assessed the accuracy of curettage and articular involvement by comparing pre- and post-operative CT images.

In all cases, deeply seated or multi-cystic tumors were sufficiently removed according to the pre- and post-operative fusion CT images. The subchondral bone was punctured when the initial thickness of the subchondral bone was less than 2.5 mm. However, use of the computer-guided burr was safe if the thickness of the subchondral bone was greater than 3 mm.

Computer-assisted curettage is a safe and useful method for localizing deeply seated benign bone tumors. However, use of the burr should be avoided when the bone thickness is less than 3 mm to avoid major tissue damage.

• Navigation
• bone tumor
• tumor resection
• computer assisted surgery
• image fusion

Introduction

Intralesional curettage is an effective surgical option for patients with benign bone tumors, enabling functional damage to be minimized. It has been reported that the overall recurrence rate of benign tumors following curettage is less than 10% [1], [2]. However, giant cell tumors or chondroblastomas require more careful management due to their high rate of recurrence [3], [4]. Thus, local adjuvants should be considered to decrease the local recurrence rate in these cases. Adjuvant treatments include cryosurgery, chemical or thermal adjuvants, or the use of a high-speed motorized burr, all of which are believed to remove microscopic deposits of tumor remaining in the tumor cavity [5]. Creation of a large window in the nearby cortex allows for more accurate visualization of the deep area of the tumor, but it is still difficult to see the fragile tissues over the sclerotic wall of the cavity. An accessory cavity of the tumor, major neurovascular structures, or joint cartilage may be located over the sclerotic wall and cannot be visualized even with a large cortical window. In such areas, aggressive curettage or burring is often avoided for safety reasons. In addition, creation of a large window is impossible in some anatomical locations, such as the femoral head, the talus, and the center of the epiphysis of other bones, without performing an extensive joint dislocation.

Computer-assisted surgery has demonstrated outstanding outcomes in recent reports. Wong et al. [6] performed 13 consecutive tumor resection surgeries with navigation guidance using a computer-assisted fusion image. On histological examination, all resected specimens showed tumor-free margins. Cho et al. [7] performed three cases of navigation-assisted surgery for malignant bone tumors and reported that limb salvage and joint preservation were feasible. So et al. [8] reported their experience with 15 navigated procedures for bone tumors of the extremities and pelvis, showing that tumor-free margins were achieved as planned pre-operatively. Computer-assisted surgery has thus become one of the more important assistance tools for malignant bone tumor surgery.

In this context, we considered that computer-assisted surgery could be applied to benign bone tumors to localize a deeply seated tumor or an accessory cavity over the sclerotic bone and to enable real-time monitoring of the burr tip to prevent major tissue damage.

Recently, several authors have reported that computer-assisted surgery was successfully applied to benign bone tumors such as osteoid osteomas in the spine and fibrous dysplasia of the jaw. Osteoid osteomas located close to the spinal cord were safely removed without nerve damage and pain was relieved in all patients [9], [10]. The removal of a large fibrous dysplasia in complex facial bone was also greatly assisted by computer-aided navigation [11]. However, there is no report of computer-assisted surgery being applied to tumor removal in the lower limb, especially near the joint.

We hypothesized that computer-assisted curettage would be a useful tool for localizing deeply seated tumor cavities and for avoiding damage to major tissues close to the cavity wall, such as articular cartilage and neurovascular tissues, while burring. In this case series study, we evaluated the accuracy and limitations of computer-assisted surgery in patients with benign bone tumors.

Patients and methods

Study population

We prospectively designed a protocol for computer-assisted surgery which was approved by the institutional review board of Samsung Medical Center (SMC IRB No. 200808058). The eligibility of subjects for computer-assisted surgery was judged by analyzing pre-operative CT or MR images. We collected study patients consecutively according to our inclusion criteria as summarized below, and all of the patients provided written informed consent.

The inclusion criteria were as follows: (1) a benign bone tumor located deep in a bone of an extremity which was difficult to approach without damaging major structures; (2) a benign bone tumor that had a risk of joint violation during tumor removal, e.g., tumors located within 5 mm of subchondral bones on CT or MRI images; and (3) a benign bone tumor close to major neurovascular tissues which could be damaged during burring.

The exclusion criteria were as follows: (1) refusal of computer-assisted surgery after explanation of the surgical protocol; (2) medically high-risk patients who could not withstand a long operation, such as might result from the added navigation registration time; and (3) patients whose possibility of joint preservation was low due to critical involvement of the joint at initial presentation.

Patient demographics

Ninety-seven patients with benign bone tumors underwent surgical treatment at our center between June 2008 and July 2010. Ten patients were eligible for inclusion in our study; of these, two patients were excluded due to registration failure of the navigation machine upon pre-operative imaging. The registration failed in one patient because the absorbable pin inserted for the fiducial marker was accidentally removed just before registration processing. In the other patient, surface registration was impossible due to the limited exposure of the bone surface after a minimally invasive approach. Ultimately, therefore, eight patients were enrolled in this study. No patients refused the computer-assisted surgery.

The mean age of the patients at surgery was 28.8 years (range: 14-55 years). There were five male and three female patients. The locations of the tumors were the proximal femur in two patients, the distal femur in three patients, the proximal tibia in two patients, and the talus in one patient. The pathology of each tumor as confirmed by a pathologist is described in Table I. The mean follow-up duration was 13.5 months (range: 7-28 months).

Table I.  Patient demographics

Patient No.	Age/sex	Pathology*	Location	Follow-up (months)	Operation time (min)	Estimated blood loss (ml)
1	21/M	CB	Distal femur	28	238	150
2	55/F	GCT	Talus	23	154	0
3	47/M	BFH	Proximal tibia	19	133	300
4	26/M	CB	Distal femur	8	187	100
5	16/M	CB	Proximal tibia	7	156	50
6	14/F	CB	Proximal femur	9	192	100
7	28/M	GCT	Distal femur	8	130	100
8	23/F	GCT	Proximal femur	7	175	330
Average	28.8	–	–	13.5	171	141

*CB: chondroblastoma; GCT: giant cell tumor; BFH: benign fibrous histiocytoma.

Baseline study

Pre-operative MRI and/or CT images were evaluated to select the cases eligible for computer-assisted surgery. If the diagnosis was uncertain, CT-guided aspiration biopsy (three cases) or open biopsy (three cases) was performed to determine the histological diagnosis. The baseline functional scores for the Musculoskeletal Tumor Society (MSTS) and Toronto Extremity Salvage Score (TESS) were obtained for each patient [12], [13].

Navigation registration and surgical technique

The Stryker Navigation System (Stryker, Mahwah, NJ) was used for all procedures. To facilitate accuracy of the computer-assisted registration, Kirschner wires were inserted into the bone adjacent to the tumor in two patients under local anesthesia, just before they underwent CT scanning, as landmarks for the fiducial-based registration. Surface registration was used for the remaining six patients. The image-to-patient registration was performed using paired-points methods in conjunction with CT images and fiducial markers (Kirschner wires) or surface markers in order to match corresponding points in the actual anatomy and in the pre-operative CT images.

Following skin incision, a wide cortical window was created using a motorized saw in order to visualize the tumor cavity completely. If the tumor was deeply seated relative to the window, a tunnel was created using a curette or burr. Intermittent probing with the computer-guided system was attempted during curettage to localize the tumor cavity. Sharp curets were used to remove the gross tumor from the cavity wall. We used a specially designed burr which is attached to the navigation probe in order to monitor the location of the burr tip within the tumor cavity in real time (Figure 1). The high-speed motorized burr extended the bony margin a few millimeters over the cavity wall. The burr was stopped when the tip was 2-3 mm from the neurovascular bundles according to the navigation system. The cavity was then bathed in 99% ethyl alcohol for a few minutes, followed by vigorous irrigation with saline. The empty cavity was filled with either bone cement (seven cases) or allogenic cancellous bone graft (one case). After wound closure, compression dressing with temporary immobilization was performed. Post-operative care was the same as that for patients operated on with conventional methods. Briefly, a gentle active range-of-motion exercise was initiated after pain was resolved, followed by crutch or walker ambulation with protective weight-bearing.

Figure 1. The specially developed burr attached to the navigation probe in order to monitor its location within the tumor cavity.

Primary and secondary outcome measures

The accuracy and intra-operative safety of computer-assisted curettage were evaluated as primary outcomes. Post-operative CT was performed to determine the presence of a residual tumor cavity.

We compared pre-operative and post-operative CT images with matched slices using an image fusion program (iNtellect Navigation System version 1.0, Stryker). To determine the accuracy, the adequacy of the resection margin and the involvement of the articular joint were evaluated in fused images. When the surgical margins in the post-operative CT scan completely covered the tumor visible in the pre-operative CT scan for all matched sections in the three planes (axial, sagittal and coronal), it was assumed that curettage was adequate.

The articular involvement was evaluated by measuring the subchondral bone thickness in the pre- and post-operative CT images (Figure 2). Pre- and post-operative CT images were matched using the aforementioned fusion program in each of three planes (axial, coronal, and sagittal). The thickness of the subchondral bone was measured at the closest margin to the articular surface on pre- and post-operative CT images. A total of 193 CT sections from five patients (cases Nos. 1, 2, 4, 5 and 7), including axial, coronal, and sagittal images, were evaluated for pre- and post-operative subchondral bone thickness. If the thickness of the subchondral bone on the post-operative CT images in any section was less than or equal to 0 mm, we assumed that the articular cartilage had been punctured in that case.

Figure 2. Articular involvement was evaluated by measuring subchondral bone thickness on the pre- and post-operative CT images. A representative case of punctured articular cartilage during burring is shown (case No. 4). The thickness of the subchondral bone between the tumor (or cement) and the outer cortex was measured via pre- and post-operative three-plane CT images. The distance between the two arrows indicates the subchondral bone thickness for each section of the matched CT images. On pre-operative sagittal and coronal images, the subchondral bone was extremely thin like paper (less than 1 mm) (B and C). This thin subchondral bone was punctured during surgery, and bone cement invaded to approximately 1 mm beyond the articular surface (E and F).

We also evaluated some safety aspects of the computer-guided surgery procedure, including operation time, blood loss, neurovascular injury, and other unexpected events.

The secondary outcome measures were functional and oncological outcomes. Post-operative functional scores, including the MSTS score and TESS, and complications related to the surgery, were recorded at every outpatient visit. We evaluated local recurrence using CT or MRI every six months following the operation. The presence of distant metastasis in aggressive benign tumors was evaluated initially and six months after the operation by chest CT.

Results

Primary outcome measures

The accuracy of navigation-assisted curettage was determined with regard to adequate resection margin and involvement of the articular joint. An adequate resection margin was determined based on a comparison of pre- and post-operative CT images. When the two images were fused, the outer resection margin in the post-operative CT scan covered the tumor cavity in the pre-operative CT scan in three planes for all eight cases (Figure 3). Also, there was no accessory cavity remaining on post-operative CT (see Figure 4, case No. 4; a unique case showing an accessory daughter cavity over the main tumor). Accordingly, we concluded that computer-guided curettage was adequate for removing benign bone tumors.

Figure 3. Representative pre- and post-operative CT scans and fusion images. All axial and coronal pre- and post-operative images were matched using a program of the Stryker Navigation System (iNtellect Navigation System version 1.0). Pre-operative images were converted to a green color, and post-operative images were red. The two images were then merged to compare the curetted margins of the benign bone tumors.

Figure 4. An accessory cavity over a sclerotic wall was successfully treated using the navigation system. Pre-operative CT images (A to C) showed that the accessory tumor cavity was located medial to the main cavity over the sclerotic wall. In post-operative CT images (D to F), not only the accessory cavity but also the main cavity were completely filled with bone cement. We could easily localize the accessory cavity in three dimensions using the specially designed burr equipped with a navigation probe (G).

Five tumors (cases Nos. 1, 2, 4, 5 and 7) were located close to the articular cartilage (see Figure 5, representing case No.1). In those cases, puncture of the articular cartilage occurred during burring despite the use of navigation assistance. To elucidate the risk factors for articular joint involvement, the relationship between pre-operative thickness of the subchondral bone and violation of the articular surface was evaluated using pre- and post-operative CT images (Figure 2). The thickness of the subchondral bone was measured at the closest margin to the articular surface on pre- and post-operative CT images. We measured each thickness in all three planes and graphically depicted the result of one representative case (Figure 6, case No. 1). It was observed clearly that that the computer-guided burr punctured the subchondral bone when the initial thickness of the bone was less than 2.5 mm.

Figure 5. A 21-year-old man presented at our outpatient clinic with vague knee pain. A chondroblastoma was detected in the medial femoral condyle close to the articular cartilage on pre-operative CT and MRI coronal images (A and B). The articular cartilage (arrow) was intact on the pre-operative MRI image (B). Following navigation-guided curettage, the tumor was removed and the cavity filled with cement (C). The articular cartilage (arrow) was intact after surgery (D).

Figure 6. Subchondral bone thickness was measured on pre- and post-operative CT images as shown in Figure 2. The results for axial (A), sagittal (B), and coronal (C) CT image sections for case No. 1 are shown separately. All three-plane (axial, sagittal, and coronal) images were measured and graphed. Each thickness value was depicted on the graph according to the serial section number of the CT image. The horizontal axis represents consecutive section numbers of the CT image, and the vertical axis is the thickness of the subchondral bone (mm) as measured in Figure 2. The gray shaded area indicates a subchondral thickness of less than 0 mm. The dashed red horizontal line indicates the thickest pre-operative subchondral bone when the articular joint was punctured.

A total of 193 sections of the fusion images from the aforementioned five cases were analyzed. In the preoperative CT scan, 146 sections (75.6%) showed subchondral bone thickness of less than 3 mm at the thinnest point, and 42 (28.8%) of those lesions were punctured. However, use of the computer-guided burr was safe if the thickness of the subchondral bone was greater than 3 mm (Figure 7)

Figure 7. All five of the punctured articular cases are summarized in three graphs (one each for axial [A], sagittal [B], and coronal [C] images). The post-operative thickness of the subchondral bone is plotted against the pre-operative thickness. When all the measured values of the five cases were analyzed together, a strong positive correlation was observed between the pre- and post-operative subchondral bone thicknesses. We also found that a pre-operative subchondral bone thickness of 3 mm was the upper limit for puncture of the articular cartilage. The gray shaded area indicates a post-operative subchondral thickness of less than 0 mm. The dotted straight line indicates points where the subchondral thickness on pre- and post-operative CT images is identical. The black vertical line indicates the thickest pre-operative subchondral bone in which the articular joint was punctured after curettage.

We also evaluated safety aspects of the proposed technique. The mean operation time was 171 min (range: 130-238 min), and the mean estimated blood loss was 141 ml (range: 0-330). In five cases, the tumor cavity was close to neurovascular tissues. One representative case was a 47-year-old man with a benign fibrous histiocytoma in the proximal tibia (case No. 3). The mass was located in the posterior aspect of the proximal tibial epiphysis and was extremely close to the neurovascular bundle (Figure 8). Using the navigation system, we were able to avoid having the high-speed burr violate the neurovascular tissue through the paper-thin cortex and the curettage was completed safely. There was no neurovascular injury in any of the five cases.

Figure 8. A representative case of benign fibrous histiocytoma on the proximal tibia. The tumor was shown to be located very close to the popliteal artery and tibial nerve on preoperative MRI (A and B). There was no neurovascular compromise after navigation-guided curettage, and post-operative MRI showed successful integration of the allogenic bone graft (C and D). Since we used a specially designed burr attached to the navigation apparatus, it was possible to monitor the burr tip in real time, thereby avoiding damage to the neurovascular bundle (E).

Cement leakage into the joint space was observed in one case (case No. 1), being noticed during the operation. In this case, we extended the incision to the joint and completely removed the cement material. At the last outpatient follow-up for this patient, there was no complaint such as limitation of the knee joint motion or joint pain. Following this incident, the punctured area was covered with mesh or a Gore-Tex patch from the inside of the tumor cavity to prevent cement leakage, and there was no cement leakage into the joint space thereafter. There were no other unexpected events related to navigation-assisted surgery during surgery or the immediate post-operative period.

Secondary outcome measures

The average pre-operative MSTS score was 16 (53%; range: 40–70%), while the average TESS score was 68 (range: 49–85) (Table II). Postoperative MSTS and TESS scores were measured at six months, one year, and two years after surgery. The mean value of the final MSTS score was 26 (86%; range: 57–97%), while that of the TESS was 84 (range: 64–93).

Table II.  Clinical results

Patient No.	Age/sex	Pathology*	Location	Pre-op MSTS (%)	Post-op MSTS (%)	Pre-op TESS	Post-op TESS
1	21/M	CB	Distal femur	15 (50)	24 (80)	64	73
2	55/F	GCT	Talus	12 (40)	17 (57)	49	64
3	47/M	BFH	Proximal tibia	14 (47)	27 (90)	70	93
4	26/M	CB	Distal femur	20 (67)	29 (97)	85	90
5	16/M	CB	Proximal tibia	21 (70)	26 (87)	73	90
6	14/F	CB	Proximal femur	16 (53)	25 (83)	70	83
7	28/M	GCT	Distal femur	12 (40)	29 (97)	49	90
8	23/F	GCT	Proximal femur	17 (57)	29 (97)	84	90
Average	28.8	–	–	16 (53)	26 (86)	68	84

*CB: chondroblastoma; GCT: giant cell tumor; BFH: benign fibrous histiocytoma.

No local recurrence was observed on MRI or CT scans in any of the cases through the final outpatient follow-up (mean: 13.5 months; range: 7–28 months). In patients with chondroblastoma and giant cell tumor, a chest CT was evaluated to detect lung metastasis six months after surgery, and there was no distant metastasis as of the last follow-up. No other significant complications were observed during the follow-up period.

Discussion

In this study, we applied computer-assisted surgery to the curettage of benign bone tumors. Deeply seated tumor cavities or multi-cystic tumors were easily localized using computer-assisted surgery, enabling adequate tumor removal. Chondroblastomas may be among the best candidates for computer-assisted curettage as they are usually located in the epiphysis, close to the articular joint. In advanced cases, the subchondral bone had been lost and the articular cartilage was damaged prior to initial detection. In such cases, preservation of the articular joint near the tumor is quite challenging, and adequate curettage sometimes requires an arthrotomy and dislocation of the involved joint, which increases morbidity of the joint.

Since computer-assisted curettage allowed us to navigate the burr three-dimensionally in the tumor cavity, aggressive exposure was not necessary to visualize deeply seated tumors located in the femoral head (cases Nos. 6 and 8), the talus (case No. 2), or the center of the femoral condyle (cases Nos. 1, 4 and 7). For those cases, the benefits of computer-assisted surgery included a small skin wound, a short surgical time, early rehabilitation, and no complications related to arthrotomy.

Creation of a wide cortical window has been a standard technique for curettage of bone tumors. The window size should be as large as the tumor cavity in order to curette tumors in the side wall. Computer-assisted curettage techniques are no exception in this regard. One of the reasons for this requirement is that there are no reliable surgical tools for adequately removing tumors from the curved side wall of the cavity or from the inside of the near cortex. Navigation tools or a burr that can detect the curved wall of the cavity have not yet been developed, and these need to be investigated in order to minimize the size of the cortical window.

Another limitation in the current computer system was that we could not delineate the areas in which burring or curettage had already been performed. Accordingly, some parts of the bone should have been over-curetted in order to ensure sufficient removal of tumor. Finer curettage will be possible if we can develop a new computer program which marks the track of the burr on the CT image in real time.

This method is theoretically useful for preventing the burr from damaging normal tissues close to the tumor cavity because the computer-assisted system allows real-time monitoring of the burr tip. However, it is still risky to use a motorized burr in very thin bone cortex, even when guided by a computer, because the navigation error can be up to several millimeters, even when the registration error is less than 1 mm [14]. Unfortunately, in our case series, computer-assisted curettage could not prevent puncture of the articular cartilage by the burr when the thickness of the bone was less than 3 mm. However, the navigation system helped us to determine the location of the puncture site in three dimensions and to cover the lesion with a Gore-Tex patch to prevent leakage of bone cement during cement packing.

Despite this current limitation, computer-assisted surgery was still useful for detecting thin cortex close to major neurovasculature where the use of a burr should be avoided. We only used the burr on the thin cortex that was not close to major neurovasculature according to the computer: If the area was close to major neurovascular tissues, curettage was carefully performed using only a blunt curette.

Regardless of the window size, an accessory cavity may be missed if it is covered with hard sclerotic bone. Intra-operative radiographs or fluoroscopy may not localize the 3D location of the accessory tumor; however, the computer-assisted method was very useful for localizing the exact location of the accessory lesion, which could not be determined from a simple radiograph (case No. 4). The current navigation system and tools were adequate for finding the exact location of the accessory lesion, and this is the major benefit of computer-assisted surgery for benign bone tumors at present.

A pre- and post-operative fusion image was also useful for assessing the adequacy of curettage and joint involvement. In cases of bone cementing, we found that the fusion image gave better information than was obtained in bone-grafting cases. In the latter, assessment of the tumor margin with the fusion image was not feasible because the extent of curettage was not clear on the CT scan.

This study was not designed to compare the outcomes with computer-assisted surgery to those obtained using conventional methods; thus, its limitations include the small number of patients, the short-term follow-up, and the lack of a control group in order to determine the superiority of computer-assisted surgery. However, computer-assisted curettage was found to be beneficial for deeply seated epiphyseal benign tumors and relatively safe, given awareness of the paper-thin cortex. With the development of software and surgical tools for computer-assisted curettage, improved oncological and functional outcomes for patients with benign epiphyseal bone tumors are expected.

Declaration of interest: This study was supported by a Samsung Medical Center Clinical Research Development Program grant, # CRS109-39-2.