Stereotactic Body Radiation Therapy for Thoracic Malignancies: An Analysis of MR-Guided Radiation Therapy Gating Tracking Accuracy

Abstract

Stereotactic body radiation therapy (SBRT) is effective for thoracic malignancies. Yet, accurate delivery to the target can be challenging due to respiratory motion. OAR sparing is crucial to avoid acute and late toxicities, especially when delivering high SBRT doses per fraction. With MR-Image Guided Radiotherapy (MR-IGRT), lung targets can be tracked in real time on planar intra-fractional MRI and treatment can be interrupted when the target deviates far beyond its planned position. In this study, we sought to evaluate tracking performance of MR-IGRT for lung tumors. Six patients with lung cancer treated with MR-IGRT using respiratory tracking and gating were selected according to observed intra-fraction tumor motion. Cine images, tracked contours, gating signals and gating parameters were collected. For each fraction, up to 200 randomly selected frames from intra-fractional imaging were selected to sample the tumor’s motion extent during beam on and beam off. A radiation oncologist re-contoured the target on all frames. Tracked and manual contours were first compared using Dice similarity coefficient (DSC) and Hausdorff distances (HD). Gating signals were also calculated for manual and tracked contours. From these signals, fractions were determined for: a. beam-off instances where beam should have been-on and b. beam-on frames where beam should have been off. Lastly, target excursions beyond the PTV were identified on beam-on instances. Patients were treated for 3 (n=2) or 5 (n=4) fractions. The target was manually contoured in more than 4600 frames. All patients had cranial-caudal tumor motion greater than 2 cm as measured on 4DCT. Mean GTV volume was 44.4 cc (1.1-143.7). Median treatment duty-cycle (beam-on time %) was 69% (37-100). As summarized in Table 1, DSCs and HDs were systematically larger on beam-off frames than beam-on frames, possibly due to tracking difficulty when the target deviates further from its expected position. Agreement between gating signals generated both contours varied fraction-by-fraction. Compelling metrics for deviation were measured between tracked and manual contours as quantified using conventional algorithms - such as DSC and HD. However, this is a pilot study with limitations and therefore requires further dosimetric analysis and clinical correlation with patient outcome.Abstract TU_24_3553; Table 1DSCs and HDs calculated between manual and tracked contours for beam on/off frames; treatment duty-cycle.PatientBeam OffBeam OnDuty-Cycle (%)DSCHD (mm)DSCHD (mm)MedianMinimumMaximum10.71 ± 0.0718.5 ± 2.40.73 ± 0.0816.8 ± 3.6716610020.66 ± 0.0210.2 ± 2.10.66 ± 0.048.9 ± 2.46055673----0.57 ± 0.085.6 ± 0.610010010040.50 ± 0.1611.4 ± 1.80.55 ± 0.1410.2 ± 1.765537950.81 ± 0.0214.4 ± 1.10.81 ± 0.0412.3 ± 1.847375960.69 ± 0.0712.2 ± 2.40.72 ± 0.058.0 ± 1.39550100Cohort0.67 ± 0.1713.4 ± 5.30.69 ± 0.1610.8 ± 5.56937100 Open table in a new tab