4DCT-Based Correlation of Respiratory Motion of Lung Tumors with Implanted Fiducials and an External Surrogate

Abstract

To use free-breathing 4DCT data to test the hypothesis that lung tumor respiratory motion is well correlated with motion of airway-implanted markers and/or a commonly used external surrogate. In an institutional review board protocol, 24 lung cancer patients had an implant of 3 electromagnetic transponders into small airways in or near the tumor. All had 4DCT at simulation according to the signal from vertical (Y) motion of an external reflecting block (EXB). The EXB was completely imaged in 11 cases. Here we consider the transponders as passive fiducials (FIDs) and respiratory motion as determined from 4DCT. FIDs, tumor (GTV) and total EXB were contoured on each 4DCT phase. Coordinates (X=lateral, Y=vertical, Z=longitudinal) of each structure centroid and the 3 FID array centroid (CFID) were found with a planning system and represent structure motion. For each patient’s imaged breathing motion, the Pearson R-squared (Rsq) correlation in each direction was evaluated between CFID and GTV and also (11 patients) between 3D GTV and EXB Y. Linear correlation between maximum excursions in corresponding directions was evaluated for the full population for GTV vs CFID (24 pts) and 3D GTV excursion vs EXB Y for the 11 cases. • The median GTV was 6.32 cc [0.14-283.9], skewed toward small GTVs with 6 GTVs > 90 cc, remainder <35 cc. • Excursion medians and ranges (cm) are: GTV X=0.19 [0.03-0.68], Y=0.25 [0.03-1.54], Z=0.76 [0.06-1.95]; CFID X=0.15 [0.03-0.95], Y=0.29 [0.08-1.51], Z=0.53 [0.1-2.0]; EXB Y=0.45 [0.1-1.02]. • Within each patient, individual FID and CFID motions in the same direction correlated well for Y (median Rsq=0.85) and Z ( median Rsq=0.92,) but less so for X (median Rsq=0.59). • Median correlations (Rsq) between GTV and CFID centroids for individual patients are: X GTV vs X CFID 0.48 [0.03-0.88] with 8/24 Rsqs > 0.6; Y GTV vs Y CFID 0.79 [0.09-0.96] with 17/24 Rsqs > 0.6; Z GTV vs Z CFID 0.91 [0.12-0.99] with 18/24 Rsqs>0.6; 3D GTV vs Y EXB 0.81 [0.2-0.95] with 9/11 Rsqs>0.6. Correlation was significantly weaker if one or both structures had small (<0.3 cm) motion. • Linear correlation of excursions in the population was best for Y and Z GTV vs corresponding CFID direction (Rsq=0.89 for both), weaker for X (Rsq=0.59) and worst for the 3D GTV vs EXB Y of the 11 patients (Rsq=0.11). In this cohort, respiratory motion of the GTV and the 3 FID array are well correlated in longitudinal and vertical directions, less well laterally, where displacements are often small. Population-based linear correlation would allow estimation of GTV excursion from excursion of passive fiducials observed with time-varying imaging. There is good correlation of the 3D GTV motion with vertical EXB motion for individual patients but the population-based correlation of excursions for these structures is weak. Further studies including a wider range of tumor sizes are desirable.