Evaluation of the accuracy of the CyberKnife Synchrony™ Respiratory Tracking System using a plastic scintillator.

Abstract

The Synchrony™ Respiratory Tracking System of the CyberKnife® Robotic Radiosurgery System (Accuray, Inc., Sunnyvale, CA, USA) enables real-time tracking of moving targets such as lung and liver tumors during radiotherapy. Although film measurements have been used for quality assurance of the tracking system, they cannot evaluate the temporal tracking accuracy. We have developed a verification system using a plastic scintillator that can evaluate the temporal accuracy of the CyberKnife Synchrony. A phantom consisting of a U-shaped plastic frame with three fiducial markers was used. The phantom was moved on a plastic scintillator plate. To identify the phantom position on the recording video in darkness, four pieces of fluorescent tape representing the corners of a 10cm×10cm square around an 8cm×8cm window were attached to the phantom. For a stable respiration model, the phantom was moved with the fourth power of a sinusoidal wave with breathing cycles of 4, 3, and 2s and an amplitude of 1cm. To simulate irregular breathing, the respiratory cycle was varied with Gaussian random numbers. A virtual target was generated at the center of the fluorescent markers using the MultiPlan™ treatment planning system. Photon beams were irradiated using a fiducial tracking technique. In a dark room, the fluorescent light of the markers and the scintillation light of the beam position were recorded using a camera. For each video frame, a homography matrix was calculated from the four fluorescent marker positions, and the beam position derived from the scintillation light was corrected. To correct the displacement of the beam position due to oblique irradiation angles and other systematic measurement errors, offset values were derived from measurements with the phantom held stationary. The average SDs of beam position measured without phantom motion were 0.16 and 0.20mm for lateral and longitudinal directions, respectively. For the stable respiration model, the tracking errors (mean±SD) were 0.40±0.64mm, -0.07±0.79mm, and 0.45±1.14mm for breathing cycles of 4, 3, and 2s, respectively. The tracking errors showed significant linear correlation with the phantom velocity. The correlation coefficients were 0.897, 0.913, and 0.957 for breathing cycles of 4, 3, and 2s, respectively. The unstable respiration model also showed linear correlation between tracking errors and phantom velocity. The probability of tracking error incidents increased with decreasing length of the respiratory cycles. Although the tracking error incidents increased with larger variations in respiratory cycle, the effect on the cumulative probability was insignificant. For a respiratory cycle of 4s, the maximum tracking error was 1.10 and 1.43mm at the probability of 10% and 5%, respectively. Large tracking errors were observed when there was phase shift between the tumor and the LED marker. This technique allows evaluation of the motion tracking accuracy of the Synchrony™ system over time by measurement of the photon beam. The velocity of the target and phase shift have significant effects on accuracy.