Clinical log data analysis for assessing the accuracy of the CyberKnife fiducial-free lung tumor tracking system

Abstract

Purpose: The CyberKnife Xsight Lung Tracking (XLT) and 1-View tracking systems can synchronize beam targeting to a visible lung tumor with respiratory motion during irradiation without requiring internal fiducial markers. The systems use a correlation model that relates external marker positions to tumor positions as well as a prediction model that predicts the target's future position. In this study, the correlation and prediction model uncertainties related to the CyberKnife fiducial-free tumor tracking system were evaluated using clinical log data.

Methods and materials: Data from 211 fractions in 42 patients with lung tumors were analyzed. Log files produced by the CyberKnife Synchrony system were acquired after each treatment; the mean correlation and prediction errors for each patient were calculated. Additionally, we examined the tracking tumor-related parameters and analyzed the relationships between the model errors and tracking tumor-related parameters.

Results: The overall means ± standard deviations (SDs) of the correlation errors were 0.70 ± 0.43 mm, 0.36 ± 0.16 mm, 0.44 ± 0.22 mm, and 0.95 ± 0.43 mm for the superoinferior (SI), left-right (LR), anteroposterior (AP), and radial directions, respectively. The overall means ± SDs of the prediction errors were 0.13 ± 0.11 mm, 0.03 ± 0.02 mm, 0.03 ± 0.02 mm, and 0.14 ± 0.11 mm for the SI, LR, AP, and radial directions, respectively. There were no significant differences in these errors between the XLT and 1-View tracking methods. The tumor motion amplitude was moderately associated with the correlation error and strongly related to the prediction error in the SI and radial directions.

Conclusions: Clinical log data analysis can be used to determine the necessary margin sizes in treatment plans to compensate for correlation and prediction errors in the CyberKnife fiducial-free lung tumor tracking system. The tumor motion amplitude may facilitate margin determination.