A potential means of improving the evaluation of deformity corrections with Taylor Spatial Frames over time by using volumetric imaging: Preliminary results

Abstract

Objective: In this study we explore the possibility of accurately and cost-effectively monitoring tibial deformation induced by Taylor Spatial Frames (TSFs), using time-separated computed tomography (CT) scans and a volume fusion technique to determine tibial rotation and translation.

Materials and Methods: Serial CT examinations (designated CT-A and CT-B, separated by a time interval of several months) of two patients were investigated using a previously described and validated volume fusion technique, in which user-defined landmarks drive the 3D registration of the two CT volumes. Both patients had undergone dual osteotomies to correct for tibial length and rotational deformity. For each registration, 10 or more landmarks were selected, and the quality of the fused volume was assessed both quantitatively and via 2D and 3D visualization tools. First, the proximal frame segment and tibia in CT-A and CT-B were brought into alignment (registered) by selecting landmarks on the frame and/or tibia. In the resulting “fused” volume, the proximal frame segment and tibia from CT-A and CT-B were aligned, while the distal frame segment and tibia from CT-A and CT-B were likely not aligned as a result of tibial deformation or frame adjustment having occurred between the CT scans. Using the proximal fused volume, the distal frame segment and tibia were then registered by selecting landmarks on the frame and/or tibia. The difference between the centroids of the final distal landmarks was used to evaluate the lengthening of the tibia, and the Euler angles from the registration were used to evaluate the rotation.

Results: Both the frame and bone could be effectively registered (based on visual interpretation). Movement between the proximal frame and proximal bone could be visualized in both cases. The spatial effect on the tibia could be both visually assessed and measured: 34 mm, 10° in one case; 5 mm, 1° in the other.

Conclusion: This retrospective analysis of spatial correction of the tibia using Taylor Spatial Frames shows that CT offers an interesting potential means of quantitatively monitoring the patient's treatment. Compared with traditional techniques, modern CT scans in conjunction with image processing provide a high-resolution, spatially correct, and three-dimensional measurement system which can be used to quickly and easily assess the patient's treatment at low cost to the patient and hospital.

• Taylor Spatial Frame
• CT volume fusion
• movement evaluation

Introduction

The use of Ilizarov-derived Taylor Spatial Frames (TSFs), as well as similar devices, for treating severe leg injuries has become widespread [1–4]. Evaluation of various frame systems [5–7] in terms of patient outcomes [8] has been previously reported.

A leg fixation system is used when an injury or disease results in a bony deformity of the lower limb. The TSF consists of rigid segments, each of which is anchored to the bone with screws and pins, and remains in place for several months. Typically, the top part of the frame is anchored to the proximal part of the tibia, the middle part is anchored below an osteotomy, and the bottom part is anchored to the distal tibia near the ankle. Ideally, the frames attached to the bone proximally and distally to the osteotomy should function as rigid bodies.

The rods and screws which connect the frames are adjusted over time. Planar radiographs acquired at fixed intervals are used by the surgeon as input to a computer program which is part of the TSF system. Elements of the frame and the intended center of rotation are marked, with the goal being to calculate a set of adjustments to the device that will achieve the desired result: a healed bone with improved anatomy. The calculated adjustments (which may sometimes be quite radical) are carried out by twisting a series of dials on the frame that adjust the rings and struts independently of one another. These adjustments allow the surgeon to influence the healing process in three dimensions. Unfortunately, the planar radiographs lack the third dimension, forcing the clinician to envision the leg and frame in three-dimensional (3D) space, and limiting his ability to accurately quantify changes in deformation and compute the optimal adjustments.

In this study, we explore the possibility of accurately and cost-effectively monitoring the deformation of the tibia using computed tomography (CT) combined with a volume fusion technique and interactive identification of the frame and/or tibia. Specifically, we seek to monitor the relative movement (i.e., rotation and translation) between the segments of the tibia and the segments of the frame connected to the bone proximally or distally to the osteotomy, and to determine the spatial effect on the tibia over time.

Materials and methods

Serial CT examinations of two patients were investigated retrospectively using a previously described and validated volume fusion technique [9–11]. CT scans were performed with a LightSpeed QX/I helical CT scanner (GE Medical Systems, Milwaukee, WI), and the volumes were acquired with 1.25-mm collimation and a pitch of 3 (0.75 mm/rotation), at 250 mA and 120 kV, from the knee to the bottom of the foot. The volumes were reconstructed with an x-y pixel size of 0.5-0.7 mm and a slice thickness of 0.5-0.6 mm. The image matrix was 512 × 512 pixels, with a range of 1101 to 783 slices. The original CT studies were transferred in DICOM format to a computer workstation (dual Xeon 2.8-GHz PC with 6 GB memory, 120 GB of disk, running Red Hat 9.0; New Technology Solutions, Inc., Lowell, MA) via the hospital network. The CT volumes ranged in size from 578 to 411 MB and thus put more demands on the computer memory than the usual planar X-ray images used with the standard 2D TSF applications. This study was approved by the New York University Institutional Review Board.

For patient 1 (male, 38 years old), the two CT examinations, CT-A and CT-B, were separated by an interval of 33 weeks. For the first osteotomy, performed near the top of the tibia to lengthen the leg, one full and one half frame were rigidly attached to the top of the bone (near the knee), while a third frame was attached toward the middle of the tibia. Both top frames were used in registering the top frames and proximal tibia. A second, more distal osteotomy was performed 24 weeks after the first to correct a rotational deformity down toward the ankle. At this time, the pins for the top and middle frame were moved and a fourth frame applied near the ankle below the second osteotomy. Thus, the rotational deformity was straightened between the bottom and middle parts of the frame.

Patient 2 (male, 56 years old) was much more typical, as there was a single frame attached to each of the top, middle and bottom portions of the tibia. Two osteotomies were performed simultaneously; one toward the upper part of the tibia to foster lengthening, and one toward the bottom to correct for a rotational deformity. There were two sets of CT scans available, separated by an interval of 5 weeks during which no pins were surgically moved. It could therefore be ideally assumed that the frame was stable.

For patient 1, it was not possible to use the frames for registration, as the frames and pins had been moved by the surgeons between CT-A and CT-B. Instead, bony landmarks on the tibia itself were used for registration. The proximal tibia in CT-B was registered to the proximal tibia in CT-A, and the fused volume (FVOL-1) was saved with the proximal tibia now aligned. The distal tibia below the second osteotomy from CT-A was then registered with the same distal tibia region in the fused volume FVOL-1 and the resulting volume (FVOL-2) was saved. This brought both the proximal and distal tibia into alignment, and the amount of translation and rotation required provided the measure of tibial deformation.

For patient 2, the proximal frame segment fixed to the proximal tibia in CT-B was registered to that in CT-A, and the resulting fused volume was saved (FVOL-1). The proximal frame segments were now aligned. Next, the distal frame segment from CT-A was registered to that in FVOL-1, and this fused volume was saved (FVOL-2). The distal frame segments were now aligned. The proximal tibia from CT-A was then registered to that in CT-B using the pins and bony landmarks, and this fused volume was saved (FVOL-3). The proximal tibia was now aligned. Next, the volume with the proximal tibia in alignment (FVOL-3) was used in conjunction with the distal frame of CT-A to register the distal tibia, and this fused volume was saved (FVOL-4). This volume (FVOL-4) and CT-A were then used with only bony landmarks to refine the registration of the distal tibia, and this final volume (FVOL-5) was saved.

For each registration, a minimum of 10 landmarks were chosen with a sphere of adjustable radius (3.5 mm was used for the frame, which approximated the frame openings, and 1.0 mm was used for marking the tip of the frame pins and for the bony landmarks). After the final registration, the centroid (weight point) of the distal tibial landmarks was calculated and placed on the distal tibia to demonstrate the validity of the registration and to calculate the amount of lengthening of the tibia. The Euler angles, generated by the registration, gave the rotational component. The CT volumes were viewed in two and three dimensions, and the movement between the frame and the proximal bone, and the spatial effect on the tibia of the lengthening procedure, was confirmed.

A flow chart of the ideal progression of the algorithm (as illustrated fully in the case of patient 2) is shown as Figure 1.

Figure 1. Flow chart showing the procedure followed to obtain the final numerical measurement. The results presented in this paper illustrate the general algorithm (especially in the case of patient 2) and prove that each step can be performed successfully.

Results

For patient one, we were initially unaware that the top frame and pins had been moved between the acquisition of CT-A and CT-B. Hence, the top frame was initially registered using spherical landmarks of approximately the same radius as the frame holes (3.5 mm). The frame was nicely aligned, as can be seen from Figure 2a, but the proximal tibia was not.

Thereafter, the proximal and distal tibia were registered using only bony spherical landmarks 1.0 mm in radius. The alignment was quite good, as can be seen from Figure 2b. The weight points of the final set of landmarks were then calculated, and the registration is shown in two dimensions in Figure 3. The elongation of the tibia was approximately 34 mm and the rotation was approximately 10°.

Figure 2. (a) The top frames (white arrows) for patient 1 are nicely aligned (green and yellow frame surfaces coincide), but the proximal tibia (magenta arrows) is not aligned (yellow: original CT-A volume from first scan date; green: original CT-B volume from second scan date after registration–FVOL-1). The top frame and pins were moved between the first (CT-A) and second (CT-B) scans. (b) The distal tibia is now well aligned (white arrow). Bony landmarks were used on the original CT-A and CT-B volumes to register the proximal tibia, and the resulting fused volume (FVOL-2) was used together with bony landmarks on the original CT-A volume to register the distal tibia, resulting in FVOL-3 (yellow: FVOL-3; green: original CT-A volume from first scan date).

Figure 3. Weight points for registered distal tibia of patient 1. (a) Axial, coronal, and sagittal slices from the original CT-A volume from the first scan date. (b) Axial, coronal, and sagittal slices from the FVOL-3 volume derived from registering the distal tibia from FVOL-2 (proximal tibia registered) with the original (CT-A) distal tibia. The weight points, derived from the landmarks, are shown on each slice (white arrows). Over the period of 33 weeks, the elongation of the tibia was approximately 34 mm and the rotation was approximately 10°.

For patient 2, it was observed that although in FVOL-1 (the fusion of the top frames from CT-A and CT-B) the proximal tibia was almost aligned, the top pins had moved. This physiological shifting of the pins within the bone is a problem that cannot usually be addressed by changing the method of attachment of the frame. Fortunately, it can (often) be addressed through registration by selecting landmarks on features of the tibia itself. Figure 4a shows the 3D registration between the top frame, proximal tibia and pins, and Figures 5a and 5c show a 2D representation of this in the axial and sagittal planes. The top pins and proximal tibia from CT-A and CT-B were subsequently used for registration of the proximal tibia, resulting in a new fused volume (FVOL-3). As can be seen in three dimensions in Figure 4b and in two dimensions in Figures 5b and 5d, this resulted in the top frame becoming unaligned (but with the tibia properly aligned). With the proximal tibia in alignment, the bottom frame from CT-A and FVOL-3 was aligned and the fused volume (FVOL-4) was saved. The distal tibia was then in good alignment, as can be seen in Figure 6a. In this registration, the bottom pins were slightly off, but not nearly as much as at the top. Using this fused volume (FVOL-4) as a starting point, bony landmarks were chosen on the distal tibia from CT-A and FVOL-4, and an improved registration was obtained, as shown in Figure 6b, with the bottom frames, pins, and distal tibia well aligned (FVOL-5). The weight points of the final set of landmarks were calculated as before, and the registration is shown in 2D in Figure 7. The elongation of the tibia is approximately 5 mm and the rotation is approximately 1°.

Figure 4. (a) For patient 2, the top frame and proximal tibia are well aligned when the top frame only is used for the registration (FVOL-1) (yellow: CT-A from first scan date; green: FVOL-1). However, the proximal pins (white arrows) are clearly not aligned. (b) The proximal tibia and pins are then used to align the proximal tibia (FVOL-3). This results in the pins (white arrow) and proximal tibia being better aligned, but the top frame (magenta arrow) is no longer aligned (yellow: CT-A from first date; green: FVOL-3).

Figure 5. Left panels: Axial view (a) and sagittal view (c) showing the registered top frame for patient 2 (FVOL-1). The proximal tibia, however, is not quite in alignment, nor are the pins. Right panels: Axial view (b) and sagittal view (d) after registering the proximal tibia using landmarks only on the proximal tibia and pins (FVOL-3).

Figure 6. (a) The fused proximal volume (FVOL-3) for patient 2 was used to align the bottom frame (yellow: fused distal volume (FVOL-4); green: original CT-A volume from first scan date). Both the bottom frame and the distal tibia (white arrows) are well aligned. The bottom pins (magenta arrow) are slightly off, but not nearly as much as at the top. (b) The fused distal volume (FVOL-4) was used with bony landmarks to further align the distal tibia (FVOL-5). The distal frame, distal tibia (white arrows) and pins (magenta arrow) are all well aligned (yellow: FVOL-5; green: original CT-A volume from first scan date).

Figure 7. Weight points for registered tibia of patient 2. Column (a): Axial, coronal, and sagittal slices from original CT-A volume from first scan date. Column (b): Axial, coronal, and sagittal slices from volume derived from final registration, FVOL-5, of distal tibia. The weight points, derived from the landmarks, are shown (white arrows) on each slice. Over the period of 5 weeks, the elongation of the tibia was approximately 5 mm and the rotation was approximately 1°.

Discussion

Since the invention of the Hoffman fixation system in 1938, much work has been done on the use of various fixation systems [12], [13], especially those intended for use on the tibia [14]. More recently, a group of researchers have been specifically investigating the Ilizarov-TSF systems [15–17].

In this preliminary study, we present the possibility of evaluating over time the course of treatment of leg injuries using CT as a supplement to, and potentially a replacement for, the planar radiographs that are routinely used. The current TSF computer system is based only on 2D data, but could be extended to use 3D data.

We have previously investigated the use of serial CT examinations in total hip arthropasty [10], [18], [19] and the placement of elbow and spine [20], [21] prostheses. CT has sufficient resolution in all axes, and current reconstruction software is able to suppress the artifacts produced by the high-attenuation material composing the frame. Furthermore, modern CT units also have a relatively low effective dose. We experimentally calculated an effective dose of 0.2 mSv using 120 kV and 140.0*0.5 mAs, and 0.03 mSv for a “care dose” with the same 120 kV but 20*0.5 mAs. This should be contrasted to an effective dose of 0.01 mSv for a series of 5 X-rays (0.002 mSv each at 50 kV and 12.5 mAs or 5 mAs). These effective dose estimates are comparable to those given in the literature [22], [23]. Thus, the use of serial CTs is now ethically acceptable, given the therapeutic value of the data that can be obtained from this approach.

The use of high-resolution 3D data is advantageous, because this volumetric data can be visualized interactively from any viewpoint. With planar radiographs, the viewpoint must be decided in advance, and the information generated by the plane radiogram may not be the information needed (thus necessitating a new radiographic procedure). That is, in order to position the patient correctly in the 2D examination, the person performing the planar radiography must have a good understanding of 3D to 2D projection issues and, more importantly, the view that is likely to be of interest to the surgeon. By using CT, these issues are obviated.

Furthermore, the interpretation of planar radiographs requires good 3D mental visualization skills. As the planar images are projective images that can produce at most a stereo view, there is always a problem with overlapping structures and features and varying magnifications, which could lead to ambiguity concerning the details in a specific case. As CT produces a 3D data set, this reduces the requirement for 3D mental visualization skills and may potentially lower the interpretation time, making it possible for the surgeon to better evaluate the course of treatment. At the Karolinska Institute, the orthopedic surgeons have found the use of CT volumes to be superior to that of planar radiographs and most often use them with the TSF treatment-planning system.

It remains to be determined whether this preliminary method can be used to replace the current 2D analysis. At present, we can obtain quantitative measurements from the 3D data, and in a future prospective study we will perform this analysis using data collected over time in conjunction with the adjustment of the frame.

The advent of computer workstations supporting 8 GB or more of memory and multiple central processing units enables the clinician to deal with the very large data sets generated by the volumetric CT scans at a relatively low cost. The availability of inexpensive high-capacity disks (1 terabyte) and very high performance graphics devices has reduced the cost of the capital investment for the CT and computer equipment. This capital cost is significantly less than the time and operating costs associated with the traditional method (which represent ongoing operational costs rather than a one-time capital investment). CT data acquisition might also shorten the time required for the examination, and decrease the cost as well as influence the timeliness of the treatment.