Abstract
Primary internal fixation of uncomplicated scaphoid fractures is growing in popularity due to its advantages over conventional cast fixation. Performing the procedure percutaneously reduces the risk of infection and soft tissue damage, but can be tricky because of the small size and complex three-dimensional (3D) shape of this bone. Computer-assisted navigation has been an invaluable tool in other pin insertion procedures.
This in-vitro study aimed to evaluate two different rendering techniques for our navigation interface: (i) 3D volume rendering of the CBCT image to show digitally-reconstructed radiographs of the anatomy, and (ii) volume-slicing, analogous to CT-images.
As the shape of the scaphoid is highly variable, a plastic model of the wrist was constructed in order to provide consistency that would not be possible in a cadaver-based study. The plastic model featured a removable scaphoid such that a new one was replaced between trials. Three surgeons each performed eight trials using each of the two navigated techniques (yielding a total of 48 trials for analysis). Central placement of scaphoid fixation has been linked with mechanical stability and improved clinical outcomes, thus the surgical goal was to place a K-wire to maximise both depth from the surface and length of the drill path. The wire was drilled through the scaphoid, from distal to proximal, allowing for post-trial analysis of the drill path. A ceiling-mounted OptoTrak Certus camera (Northern Digital Inc., Canada) and a floor-mounted isocentric 3D CBCT C-arm (Innova 4100, GE Healthcare, France) permitted a registration transformation between the tracking and imaging systems to be computed preoperatively, before each trial, using a custom calibration device. Optical local coordinate reference bodies were attached to the wrist model and a custom drill guide for tracking with the Certus camera. During each trial, a 3D spin image of the wrist model was acquired, and rendered according to the technique under study.
For 3D volume rendering, the spin image was rendered as a digitally-reconstructed radiograph (DRR) that could be rotated in three dimensions. In the planning phase, the surgeon positioned a desired drill path on the images. Anterior-posterior and lateral views of the 3D volume rendering were used for navigation during the drilling phase. The real-time orientation of the drill guide was shown relative to these images and the plan on an overhead.
For volume-sliced (VS) navigation, the spin image was volume-rendered and sliced along the principal planes (axial, coronal, sagittal) for planning. A slider interface allowed the surgeon to scroll through the slices in each of the planes, as if they were looking at individual CT slices. Once the desired drill path was positioned, the volume-sliced views were reconfigured to show slices along the oblique planes of the planned path for navigation.
Following all trials, model scaphoids with wire intact were imaged using CT with a slice thickness of 0.625 mm. The CT series were segmented and used to construct 3D digital models of the wire and drilled scaphoid. Algorithms were developed to determine the minimum distance from the centerline of the wire and the scaphoid surface, and to compute the length of the drill path. Screw breach should be avoided as it disrupts the articular surface and may lead to a sequela of cartilage deterioration and osteoarthritic changes. The shortest distance measure was extrapolated to assess whether a standard fixation screw (Accutrak Mini, 1.78 mm radius) would have breached the scaphoid surface. There were three screw breaches noted in the 3D DRR trials, while only one occurred using volume-slicing. The minimum distance from the centerline of the wire to the scaphoid surface can also be thought of as a “safe zone” for screw breach. Although no difference in the mean distance (μ) was noted between groups (μDRR = 2.3 mm, μVS = 2.2 mm), the standard deviation (σ) was significantly higher for the DRR trials (σDRR = 0.50 mm, σVS = 0.37 mm, p < 0.1), suggesting a higher reliability of central placement using VS for navigation. In contrast, the length of the drill paths were significantly longer for the DRR trials (μ = 28.7 mm, σ = 0.66 mm) than for VS-navigation (μ = 28.3 mm, σ = 0.62 mm) at p < 0.1.
The surgical goal was to pick a path that maximised both the length of the path, as well as the minimum distance from the scaphoid surface. Algorithms were developed to find the paths that would maximise: (i) the length and (ii) the distance from the surface of the model scaphoid used in this study. The maximum possible length was 29.8mm (with a minimum distance of 2.2mm from the scaphoid surface), and the maximum distance was 3.3mm (with a length of 27.5mm). Therefore, the set of optimal drill paths had length > 27.5 mm, and distance > 2.8 mm. Of the DRR-navigated trials, 11 were below the minimum optimal depth, and only one trial was below the optimal length; 13 of the 24 trials (54%) were of both optimal length and depth. Of the VS-navigated trials, nine were below the minimal optimal distance, and four were below the minimum optimal length; 11 out of 24 trials (46%) were within both the optimal length and depth.
From this comparative study, we conclude that VS-navigation was superior in locating a central location for the fixation wire, while DRRs were superior in maximising the depth of the drill path. Thus, we propose a hybrid interface, incorporating both volume-slicing and DRRs, in order to maximise the effectiveness of navigation for percutaneous scaphoid pinning.
