An Off-Line 4D Cone Beam CT Based Correction Protocol for Lung Tumor Motion

Abstract

Purpose/Objective: Respiratory motion induces geometrical uncertainties during planning and treatment of lung cancer. An essential part of four-dimensional image guided RT (4D-IGRT) is the validation of the tumor position and trajectory during the treatment course. For that purpose we have developed a 4D-cone beam CT (4D-CBCT) system that can be used to validate tumor position and motion just prior to irradiation. The purpose of this study is to quantify the achievable margin reduction using this system in combination with an off-line correction protocol. Materials/Methods: Corrections for tumor position are initially limited to the cranial-caudal (CC) axis, which is reasonable as respiratory motion dominates in this direction. Because corrections for tumor position may compromise organs at risk (OAR), it is verified at the treatment planning stage that shifts of 1 cm in the cranial or caudal direction of the isocenter do not violate dose constraints of the OARs. 4D-CBCT scans are acquired just prior to irradiation and registered to the (3D) planning CT (pCT) on the bony anatomy. This registration is adapted manually in the CC-direction such that the PTV contours (imported from the planning system) symmetrically encompass the moving tumor as visible in the 4D-CBCT. If this extra shift in the CC-direction exceeds 1 cm, the treatment plan is re-evaluated. Finally, the setup correction is determined using a shrinking action level protocol based on the combined setup error of bone (in the left-right (LR) and anterior-posterior (AP) direction) and tumor (in the CC direction). To test this protocol, we used 4D-CBCT scans acquired on the linac of 10 lung cancer patients for ∼10 fractions per patient. These scans were first matched on bony anatomy on the pCT and subsequently the tumor motion was quantified by automatically registering a region of interest (an expansion of the delineated GTV) of the pCT with each phase of the 4D-CBCT. This procedure allows determination of amplitude and baseline of the respiratory induced tumor motion. Intra-observer variability of the manual match procedure was assessed using 3 observers. The correction protocol was simulated based on the initial and residual uncertainties. Results: Substantial systematic errors in tumor position were observed, which were mainly induced because a free-breathing CT was used for planning. The SDs of the systematic error were 1.5 (LR), 3.8 (CC) and 3.0 (AP) mm. Random variations were observed with SDs of 1.6 (LR), 1.7 (CC), and 2.3 (AP) mm. These values mainly represent variation in base-line breathing level, although in some cases asymmetric tumor regression was observed that caused a time trend. The observer variation in the manual CC-match was 1.7 mm SD. The (uncorrected) systematic setup errors of the bony anatomy were 2.5 (LR), 4.5 (CC), and 1.8 (AP) mm. Simulation of the correction protocol yielded a residual systematic deviation between the planned and actual mean tumor position of 1.9 (LR), 1.5 (CC), and 3.2 (AP) mm. Without 4D-CBCT based corrections the deviation in the CC-direction would have been 4.0 mm SD. For a tumor with a peak-to-peak motion of 15 mm, the margin in the CC direction could be reduced from 15 mm to 8 mm. Conclusions: Substantial systematic and random variations in mean tumor position occur. The proposed off-line correction protocol based on manual tumor registration in the CC direction on 4D-CBCT data reduces this uncertainty for the CC direction considerably. Since April 2005, this protocol is in clinical use in our institution. However, also large deviations in the AP direction occur that cannot yet be corrected. In the future, we will implement online dosimetric evaluation in our IGRT system, allowing safe correction for tumor position changes in all 3 directions.