The study protocol was approved by the Institutional Review Board (No. 18–19, T2019-0178). This was a retrospective comparative study (Level of Evidence III). It has been registered as NCT03764501 at ClinicalTrials.gov (registered 21 May, 2018, retrospectively registered). This study was performed in accordance with the relevant guidelines and regulations. Informed consent was obtained from all individual participants included in the study. Ten wrists from 10 distal radius malunion patients who underwent corrective osteotomy (six females, four males, mean age 59.4 years, age range 40–78) were evaluated. The image fusion group underwent 3D preoperative planning and performed corrective osteotomy with an image fusion system (n = 5). The control group was enrolled from patients who underwent corrective osteotomy using only 3D preoperative planning (n = 5).
Preoperative planning
In both groups, 3D digital preoperative planning and a surgical simulation were performed prior to surgery. Corrections to and the placement of implants were simulated using software developed by one of the authors (Zed-Trauma, LEXI Co., Ltd., Tokyo, Japan). Computed tomography (CT) of the affected and unaffected wrists using contiguous images with a slice thickness of 1–1.5 mm were taken for the simulation. Images were taken approximately 12 cm proximal to the radial joint surface. A total of 80–120 axial CT images were used for the simulation. After importing DICOM images into the software, a 3D image of the distal radius was made for both wrists. In dorsally angulated malunion, the placement of a volar locking plate was initially simulated. Computer-aided design models of different-sized implants are installed in the software; a placement image of the plate was created by calculating the correction angle required to restore volar tilt, radial inclination, and rotational deformation (Fig. 3). Stellar II locking plates (HOYA Technosurgical, Inc., Tokyo, Japan) were used in the present study. This plate system has small, medium, and large widths as well as short and long plate lengths. The plate size was selected to cover the distal radius maximally and not exceed the width of the distal radius. In addition, a sufficient length plate for the insertion of at least three screws into the radius shaft after reduction was chosen. The lengths for the distal screws were selected and a contour extraction image of the initial plate placement was saved for image fusion. The osteotomy line was set at a position that did not interfere with distal screw holes. After the osteotomy simulation, the plate and distal fragment were grouped, and the distal fragment was repositioned by adapting the proximal side of the volar locking plate to the radius shaft (Fig. 4). After repositioning the fragments, the 3D bone shape was compared with a mirror image of the unaffected radius. In the next step, simulations of the screw choices were performed for the proximal screw holes and the screw lengths for each screw hole were selected. This final reduction and implant placement image was also saved for image fusion. To compare planned and postoperative reduction shapes, preoperative plan and postoperative 3D images were created for all patients.
Image fusion system and surgical intervention
Regarding image fusion, 3D images of preoperative plans were converted to digitally reconstructed radiographs. Bone and implant contour extraction images were created for anterior–posterior and lateral views based on 3D images. Fusion images were displayed on a monitor overlapping the outline of the 3D preoperative plan and the fluoroscopic image. Corrective osteotomy was performed under general anesthesia. In the image fusion group, the outline of the planned image was displayed on a monitor overlapping the fluoroscopy image during surgery. Before starting surgery, the contour extraction image size was calibrated by measuring a known length. A surgeon performed corrective osteotomy based on the fusion image. Before osteotomy, a plate placement image was displayed on the monitor (Fig. 5). In the first step, the plate was placed on the distal radius according to the outline of the plate image and fixed with two temporary fixing wires. Outlines of the anterior–posterior and lateral views were used. According to the direction of the fluoroscopic image, the direction of the contour image was changed to the anterior–posterior or lateral view. Plate placement was checked with a fusion image for each direction. After determining the plate position at the distal radius, distal screw holes were pre-drilled. The plate was then removed leaving the temporary fixing wires. In the second step, osteotomy was performed at a level that did not interfere with the distal screws of the plate. In the third step, the plate was returned to the originally selected position under the guide of the temporary fixing wires, and the distal screws were inserted into the pre-drilled holes. Finally, the distal fragment was repositioned by adapting the proximal side of the volar locking plate to the radius shaft, and the plate was fixed with screws (Fig. 6).
In the control group, the surgeon performed reduction and placement of the plate while comparing separate images of the 3D preoperative plan and fluoroscopy during surgery. In both groups, a beta-tricalcium phosphate-based artificial bone or autologous bone graft was performed depending on the size of the bone defect. Surgeries were performed by one hand surgeon.
Evaluations
Pre- and post-operative 3D images of the distal radius were analyzed using image analysis software (BoneSimulater, Orthree, Osaka, Japan). DICOM data of CT images were imported into the software. A 3D surface model of the radius was constructed with a surface construction algorithm. The long axis of the radius was calculated from the 3D surface model of the intact part of the preoperative distal radius image. Image registration for the preoperative plan and postoperative reduction were performed using the intact part of the distal radius image. The y-axis was defined as the long axis of the radius, and the proximal direction was defined as positive. The z-axis was parallel to the orthogonal projection of a line initiating at the base of the distal ulnar sigmoid notch and continuing to the radial styloid process on a plane perpendicular to the y-axis. The radial direction on the z-axis was defined as positive. The x-axis was normal to the yz plane and the palmar direction was defined as positive. The yz, xy, and xz planes were defined as the coronal, sagittal, and axial planes, respectively. The origin of coordinates was defined as the intersection of the joint surface and the radius of the long axis on the preoperative plan image. Three reference points: (1) the radial styloid process, (2) the sigmoid notch volar edge, and (3) the sigmoid notch dorsal edge, were marked on pre- and post-operative 3D images (Fig. 7). The 3D coordinates of each reference point and the barycentric coordinates of the plane connecting the three reference points were evaluated using the 3D images of the preoperative plan and postoperative reduction.
In the sagittal view, the angle between a connecting line from reference point (2) to reference point (3) and a line perpendicular to the longitudinal axis of the radius was measured as volar tilt (3DVT). In the coronal view, the angle between a line from reference point (1) to reference point (2) and a line perpendicular to the longitudinal axis of the radius was measured as radial inclination (3DRI).
In evaluations of clinical outcomes, Mayo wrist scores [27] were recorded 3 and 6 months after surgery.
Statistical analysis
Results are expressed as the mean ± standard deviation. Distances between the preoperative plan and postoperative reduction for each reference point were measured for both groups. Differences between the preoperative plan and postoperative reduction for 3DRI and 3DVT were measured for both groups. The Shapiro–Wilk test was used to test the normality of datasets. The distances of reference points and differences in angles were compared using Welch’s t-test. The Mann–Whitney U test was used for unevenly distributed datasets (3DVT). P values of less than 0.05 were considered to be significant. All analyses were performed using BellCurve for Excel version 2.12 (SSRI Co., Tokyo, Japan).