Physics Contribution
Clinical Accuracy of the Respiratory Tumor Tracking System of the CyberKnife: Assessment by Analysis of Log Files

https://doi.org/10.1016/j.ijrobp.2008.12.041Get rights and content

Purpose

To quantify the clinical accuracy of the respiratory motion tracking system of the CyberKnife treatment device.

Methods and Materials

Data in log files of 44 lung cancer patients treated with tumor tracking were analyzed. Errors in the correlation model, which relates the internal target motion with the external breathing motion, were quantified. The correlation model error was compared with the geometric error obtained when no respiratory tracking was used. Errors in the prediction method were calculated by subtracting the predicted position from the actual measured position after 192.5 ms (the time lag to prediction in our current system). The prediction error was also measured for a time lag of 115 ms and a new prediction method.

Results

The mean correlation model errors were less than 0.3 mm. Standard deviations describing intrafraction variations around the whole-fraction mean error were 0.2 to 1.9 mm for cranio–caudal, 0.1 to 1.9 mm for left–right, and 0.2 to 2.5 mm for anterior–posterior directions. Without the use of respiratory tracking, these variations would have been 0.2 to 8.1 mm, 0.2 to 5.5 mm, and 0.2 to 4.4 mm. The overall mean prediction error was small (0.0 ± 0.0 mm) for all directions. The intrafraction standard deviation ranged from 0.0 to 2.9 mm for a time delay of 192.5 ms but was halved by using the new prediction method.

Conclusions

Analyses of the log files of real clinical cases have shown that the geometric error caused by respiratory motion is substantially reduced by the application of respiratory motion tracking.

Introduction

A wide variety of methods have been developed to manage respiratory motion in radiotherapy (1). Examples are motion-encompassing methods such as the use of an internal target volume or safety margin. Other methods include breath-hold and abdominal compression, and respiratory gating by either an external signal or internal markers. In-room respiration-correlated cone-beam CT scanning makes it possible to localize the motion trajectory just before each fraction and to adjust for interfractional baseline shifts (2).

At our center, we use the method of real-time tumor tracking. Real-time tumor tracking is performed with the Synchrony® Respiratory Tracking System, which is part of the CyberKnife® Robotic Radiosurgery System (Accuray Inc., Sunnyvale, CA) 3, 4 Briefly, Synchrony combines X-ray imaging of internal markers with a continuously updated external breathing signal. A correlation model that relates the external breathing signal with the motion of the internal markers provides a real-time update of the beam position. The correlation model is built just before the start of each treatment fraction. It is updated during the fraction by taking, every 1 to 6 min, an X-ray image pair and is rebuilt if necessary. The beam cannot be positioned instantaneously. Data processing, communication to the robotic controller, and the inertia of the robotic manipulator and the linear accelerator causes a time delay of 192.5 ms in our version of the system. A prediction method that applies adaptive filtering compensates for this delay (3).

The correlation model method and prediction method is subject to a finite accuracy. Phantom experiments have proven that Synchrony enables real-time respiratory motion tracking with an accuracy of 0.7 ± 0.3 mm.3 It is unknown to what extent this accuracy can be achieved in real patient cases. Irregular breathing, varying phase relationships between internal and external markers, and rapid baseline shifts might reduce the clinical accuracy of respiratory motion tracking. Seppenwoolde et al. quantified the residual correlation model error in patients for whom synchronized recordings were available for internal and external markers (4). As these patients were not treated with CyberKnife, tumor tracking was simulated. The average length of these recordings was 2 min with a maximum of 4 min. The same motion data were used in a dosimetric validation study using Gafchromic EBT film in a breathing phantom (5). A CyberKnife treatment fraction takes much longer than 2 min. In our current practice, the median overall treatment time to deliver a fraction dose of 20 Gy using respiratory motion tracking is 103 min (with a dose rate of 400 MU/min) (6). Although the results in both studies provide useful data on the tracking accuracy on the short time scale, the data are not conclusive on the clinical accuracy for patients who are actually treated with Synchrony.

In this study we analyzed Synchrony treatment log files of 44 patients, which included 158 treatment fractions. From the data in the files, we derived the residual error of the correlation model and the intrafraction error had Synchrony not been used. Furthermore, we analyzed the residual error that is associated with the prediction algorithm. These data can be used to contribute to the determination of proper safety margins for this method of management of respiratory motion.

Section snippets

Patient characteristics

Synchrony treatment log files of 44 consecutive patients treated for a total of 49 tumors were analyzed. Patients were treated for early-stage lung cancer or a solitary metastasis in the lung. A total of 28 tumors were treated with three fractions of 20 Gy, 16 with three fractions of 15 Gy, four with five fractions of 9 Gy, and one with six fractions of 8 Gy. Five tumors were located centrally and the others peripherally. The dose to the planning target volume (PTV) was prescribed to the 70% to

Correlation model error

The correlation model errors were small, with overall mean values (Ms) and standard deviations (s) for the 158 fractions being −0.1 ± 0.3 mm, 0.0 ± 0.3 mm, and −0.2 ± 0.3 mm, for the CC, LR, and AP components, respectively.

Figure 2 shows the distribution of the standard deviations describing intrafraction variations around the fraction mean error. The distribution of the correlation model error (σspk) is shown together with the error distribution had Synchrony not been used (σpk). The range of

Discussion

This study has shown that the residual error with Synchrony real-time respiration tracking was small in clinical practice. For a range of respiratory motion amplitudes up to 2 cm the intrafraction error was less than 2.5 mm (Fig. 2, Fig. 3). These results are comparable with the Synchrony simulation study of Seppenwoolde et al.(4) and consistent with other studies that show a reasonable correlation between the position of internal and external markers (12). The correlation, however, can be

Conclusion

The log file analyses of real clinical cases showed that the geometric error resulting from respiratory motion was substantially reduced by the application of Synchrony. Respiratory motion tracking compensated for both the intrafractional respiratory motion and for interfraction baselines shifts. Intrafraction baselines were accounted for by frequently updating the correlation model and checking the correlation model error. Therefore, respiratory motion tracking allows a considerable reduction

Acknowledgments

The authors are grateful to Warren Kilby, Ye Sheng, and Alex Li of Accuray Inc. for their help and technical support.

References (19)

There are more references available in the full text version of this article.

Cited by (0)

Conflict of interest: none.

View full text