~60 cm from isocenter and incorporates a Pb glass window to shield the tracker from radiation damage. The reference marker comprised six faces visible from.
Title Tracker-On-C: A Novel Tracker Configuration for Image-Guided Therapy using a Mobile C-arm Authors S Reaungamornrat1, Y Otake1, A Uneri1, S Schafer2, J W Stayman2, W Zbijewski2, D J Mirota1,J Yoo1, S Nithiananthan2, A J Khanna3, R Taylor1, J H Siewerdsen1,2 1
Department of Computer Science, Johns Hopkins University, Baltimore MD Department of Biomedical Engineering, Johns Hopkins University, Baltimore MD 3 Department of Orthopaedic Surgery, Johns Hopkins University, Baltimore MD 2
Presentation Preference: Lecture/Poster Keywords: surgical navigation, cone-beam CT, image-guided intervention
Tracker-On-C: A Novel Tracker Configuration for Image-Guided Therapy using a Mobile C-arm Purpose Intraoperative imaging combined with high-precision tracking is a prevalent trend in minimally invasive surgery. In many systems (e.g., infrared/video trackers), the tracker position in the operating room introduces a variety of limitations: 1) line-of-sight obstruction can pose a frequent annoyance; 2) large distance from the surgical field (~100-300 cm) can limit geometric accuracy; and 3) for video-based trackers (with potential for video augmentation), the tracker view does not match the surgeon’s perspective. To address these limitations, we investigate a configuration (Fig. 1(a)) in which the tracker is directly mounted on a C-arm providing fluoroscopy and conebeam CT (CBCT). The configuration is hypothesized to reduce line-of-sight occlusion, improve geometric accuracy, and provide a closer match between the camera and surgeon’s perspective in video augmentation. A novel reference marker maintains registration in a dynamic reference frame. The proposed system is evaluated in terms of tracking precision and new virtual fluoroscopy and video augmentation capabilities. Methods Figure 1 shows specific implementation of a videobased tracker (Claron MicronTracker Sx60). The Carm is a prototype developed for high-quality intraoperative CBCT. The mount places the tracker at ~60 cm from isocenter and incorporates a Pb glass window to shield the tracker from radiation damage. The reference marker comprised six faces visible from 360o to define a reference coordinate system enabling tracking across a full C-arm rotation. Tracking from a dynamic coordinate system registered via the hex-face reference marker was tested using a rigid skull phantom. First, the effect of the Pb glass window on target registration error (TRE) was measured. Second, the ability to maintain registration to the hex-face marker was tested by TRE measurement over 180⁰. Finally, the TRE in the proposed configuration (60 cm tracker-to-isocenter) was compared to the conventional setup (110 cm). The rigid skull phantom included 12 divots identified in diagnostic CT and divided in two groups: 6 divots as registration fiducials and 6 divots as target points for measurement of TRE. Initial measurements were performed on an optical bench with a rotation table, followed by measurements on the actual C-arm.
Two additional capabilities were evaluated. The first was generation of digitally reconstructed radiographs (DRRs) from a perspective determined by the Trackeron-C. GPU-accelerated DRRs were generated from the perspective of a handheld tool (e.g., a tracked laser pointer “virtual flashlight”). Alternatively, DRRs were computed from the pose of the C-arm itself (denoted “virtual fluoro”), requiring incorporation of C-arm geometric calibration within the pose estimation. Each provided intuitive DRR generation within the surgeon’s hands for improved workflow and reduced intraoperative fluoroscopy time (dose). The accuracy of DRRs was measured from the distance between BB centroids in real and virtual fluoroscopy of a BB phantom. Additional capability used a video-based tracker in which registered image and planning data were overlaid on the video scene. The utility of such video augmentation was evaluated in cadaver studies guided by an expert surgeon for feedback on improvements in surgical workflow (e.g., visualizing underlying anatomy, incision points, and tool trajectory without fluoroscopy). The work is distinct from seminal work on camera-augmented C-arm in that the video is derived from the tracker and not intended for direct overlay with real fluoroscopy.
Figure 1. (a) Schematic of the Tracker-on-C and conventional tracker configurations. (b) Video-based tracker within the C-arm mount. (c) Hex-face reference marker allowing tracker registration within a dynamic reference frame. (d) An example tracked tool (surgical pointer).
Results The Tracker-on-C setup is shown in Fig. 2. The “conventional” setup placed the tracker at 110 cm from isocenter, giving 1.92±0.71 mm TRE shown by the first whisker plot in Fig. 2(b). The other plots show TRE measured across 180o rotation on the benchtop, demonstrating feasibility of the hex-face reference marker to maintain registration in a dynamic reference frame, giving mean TRE of 0.70±0.32 mm. For the tracker mounted on the actual C-arm (last whisker plot), the TRE was measured across 180⁰ in 30⁰ steps, giving mean TRE of 0.86±0.43 mm. There was no statistically significant difference (p-value = 0.11) between the benchtop and C-arm measurements, and no effect on TRE due to the Pb glass window (0.87±0.25 mm with Pb glass, versus 0.83±0.25 mm without, p-value = 0.50). The improved accuracy for the proposed configuration (60 cm) was statistically significant (p-value < 0.0001). Virtual fluoro and video augmentation are illustrated in Fig. 2. Overlay of virtual fluoro (red) on a real x-ray projection of the BB phantom shows accuracy of 0.52±0.22 mm. Sub-mm accuracy was measured for Carm angulations across 180o. Video augmentation using a video-based tracker is illustrated in Fig. 2(e,f), where surgical planning data (target volume and critical structures) are overlaid on the video scene. Expert feedback highlights the potential for improved workflow and reduced dose, including C-arm setup (quickly placing the FOV), visualization of anatomy underlying the surface, quick identification of incision points, better fluoroscopy utilization (e.g., hunting for optimal oblique views), and intuitive interface to virtual DRR from any pose about the patient. Conclusion A novel configuration involving direct mounting of a tracker on a rotational C-arm was evaluated. The system demonstrated improved TRE and line-of-sight, with virtual fluoro and video augmentation capabilities providing improved workflow and reduced fluoroscopic dose. Real-time DRRs computed from the pose of tracked tools and/or the C-arm provided an intuitive interface to virtual fluoroscopic imaging. Use of a video-based tracker allowed video augmentation that improved surgical workflow in C-arm setup and visualizing instruments with respect to image and planning. Incorporation on a CBCT-capable C-arm leveraged such capabilities further in that images represented true anatomy at the time intervention, including deformation and tissue excision.
Figure 2. (a) Tracker-On-C setup. (b) TRE for the conventional setup, benchtop simulation, and Trackeron-C-arm configuration. (c) Comparison of a DRR (red) computed from Tracker-on-C pose estimation in comparison to real fluoroscopy (gray). (d) Zoomed-in view of a BB from the DRR phantom. (e) Augmented
view of surgical planning overlaid on video from the video-based tracker. (f) Zoomed-in view, including segmentations of the clivus target and critical anatomy.
