Accuracy verification of infrared marker-based dynamic tumor-tracking irradiation using the gimbaled x-ray head of the Vero4DRT (MHI-TM2000).

PURPOSE To verify the accuracy of an infrared (IR) marker-based dynamic tumor-tracking irradiation system (IR tracking) using the gimbaled x-ray head of the Vero4DRT (MHI-TM2000). METHODS The gimbaled 6-MV C-band x-ray head of the Vero4DRT can swing along the pan-and-tilt direction to track a moving target. During beam delivery, the Vero4DRT predicts the future three-dimensional (3D) target position in real time using a correlation model [four-dimensional (4D) model] between the target and IR marker motion, and then continuously transfers the corresponding tracking orientation to the gimbaled x-ray head. The 4D-modeling error (E4DM) and the positional tracking error (EP) were defined as the difference between the predicted and measured positions of the target in 4D modeling and as the difference between the tracked and measured positions of the target during irradiation, respectively. For the clinical application of IR tracking, we assessed the relationship between E4DM and EP for three 1D sinusoidal (peak-to-peak amplitude [A]: 20-40 mm, breathing period [T]: 2-4 s), five 1D phase-shifted sinusoidal (A: 20 mm, T: 4 s, phase shift [τ]: 0.2-2 s), and six 3D patient respiratory patterns. RESULTS The difference between the 95th percentile of the absolute EP (EP (95)) and the mean (μ) + two standard deviations (SD) of absolute E4DM (E4DM (μ+2SD)) was within ± 1 mm for all motion patterns. As the absolute correlation between the target and IR marker motions decreased from 1.0 to 0.1 for the 1D phase-shifted sinusoidal patterns, the E4DM (μ+2SD) and EP (95) increased linearly, from 0.4 to 3.0 mm (R = -0.98) and from 0.5 to 2.2 mm (R = -0.95), respectively. There was a strong positive correlation between E4DM (μ+2SD) and EP (95) in each direction [(lateral, craniocaudal, anteroposterior) = (0.99, 0.98, 1.00)], even for the 3D respiratory patterns; thus, EP (95) was readily estimated from E4DM (μ+2SD). CONCLUSIONS Positional tracking errors correlated strongly with 4D-modeling errors in IR tracking. Thus, the accuracy of the 4D model must be verified before treatment, and margins are required to compensate for the 4D-modeling error.