CyberKnife with integrated CT-on-rails: System description and first clinical application for pancreas SBRT

This article reports on the integration of a sliding-gantry CT-on-rails with a robotic linear accelerator.

The system consists of a SOMATOM Definition AS CT scanner (Siemens Healthcare, Forchheim, Germany) and a CyberKnife M6 FIM (Accuray, Inc., Sunnyvale, CA, USA). Additional movement programs were implemented in the robotic treatment table (RoboCouch, Accuray Inc.) to move between CT and treatment position. Acceptance testing was performed on the CT scanner according to AAPM83 guidelines, as well as safety tests for collision avoidance and electromagnetic (EM) compatibility. For the first clinical application of the system, daily dose was evaluated in five pancreas SBRT patients. A second envisioned use is the optimal alignment of the treatment beams to soft-tissue targets without the use of implanted fiducials. To this end, an offset vector feature has been implemented, which shifts the treatment center according to the daily position of the tumor relative to the spine (established by a CT scan). This offset can be applied by either moving the treatment couch (physical couch shift) or by moving the CyberKnife robot (virtual couch shift). An End-to-End (E2E) test was specifically designed to evaluate the accuracy of this feature using the Xsight Lung Tracking Phantom (Computerized Imaging Reference Systems, Inc., Norfolk, VA, USA). The position of the tumor with respect to the spine was varied by moving the insert inside the phantom and a CT scan was made for each position. The treatment plan was subsequently delivered to the phantom employing spine tracking. The test was repeated four times for a physical couch shift and four times for a virtual couch shift.

All acceptance, safety and EM compatibility testing was successful. For the first pancreas SBRT patients treated using daily CT imaging, the volume of stomach, duodenum, or small bowel receiving >35 Gy was found to increase or remain constant during treatment; however, the clinical constraint of 5 cc was not violated. For the offset vector E2E test, the reference accuracy (without any tumor shift) was (0.74, -0.61, -0.33) mm in the inferior, left, and anterior direction respectively. The difference in deviation with respect to the reference was (-0.1 ± 0.15, 0.01 ± 0.16, -0.17 ± 0.25) mm, when applying a physical couch shift. With a virtual couch shift, the deviations were (0.02 ± 0.15, 0.06 ± 0.23, -0.4 ± 0.31) mm.

The first combination of a CyberKnife treatment unit with a sliding-gantry CT scanner is operational in our department enabling future developments toward image-guided online-adaptive SBRT supported by diagnostic-quality CT imaging.