Continuous tracking of lung tumor motion at 3.3 Hz or faster is necessary to accurately represent motion

Purpose/Objective(s): Lung tumors are inherently subject to large and unpredictable motion, which make them difficult to treat with external beam radiation therapy. Missed targets can result in excessive radiation to healthy lung and other critical structures. Continuous, real-time tracking of this motion is an available solution to this problem but may not be fully necessary. We investigated the required intermittent imaging frequency necessary to accurately represent lung tumor motion during treatment. Materials/Methods: Continuous electromagnetic transponder data of lung tumor motion was collected from 45 patients (572 total fractions) across 6 institutions during external beam radiation therapy in a prospective clinical trial (NCT01396551). External intervention from the clinician, such as couch shifts, instructed breath-holds, and acquisition pauses, were manually removed from the 10 Hz tracking data according to recorded notes. The tracking data was converted to three-dimensional (3D) displacements from the initial location in each fraction. The resulting 3D data was sampled at 0.2-, 0.3-, 0.4-, and 0.5-second intervals and modeled linearly between points to represent the intermittent imaging models. Fractions lasting <60 seconds were excluded from the analysis. Sensitivity for identifying displacements greater than 3- and 5-mm throughout each patient's entire treatment was used to assess each imaging model. Imaging models were considered sufficient with ≥95% sensitivity for 3-mm displacements for ≥95% of the patients. Results: A total of 563 fractions and 6153 minutes of tracking data met inclusion criteria. Both the 0.2-second (5 Hz) and 0.3-second (3.3 Hz) imaging models were sufficiently accurate in identifying 3-mm displacements (see Table 1). The 0.4-second (2.5 Hz) and 0.5-second (2.0 Hz) imaging models failed to reliably identify 3-mm or larger displacements. Conclusion: Imaging at least every 0.3 seconds (3.3 Hz) is necessary for accurate tracking of lung tumor motion. Less frequent imaging requires a supplemental respiratory motion model.