Target localization errors from fiducial markers implanted around a lung tumor for dynamic tumor tracking

Abstract

Purpose: To assess target localization errors (TLEs) from implanted fiducial markers by three different centers of gravity (CG) and three different multiple regression analysis (MRA) approaches.

Methods: The three-dimensional (3D) positions of the markers were detected on the fluoroscopic images of 15 lung cancer patients, and the marker closest to the tumor was then assumed to be the target (Pt). The estimated target position (Pe) was calculated from three markers adjacent to the target (Pi, 1 ≤ i ≤ 3) using the equation Pe = aP1 + bP2 + cP3 + d. Pe was then calculated using three different CGs and three different MRAs. The TLE was calculated as the root-mean-square error of the difference between Pt and Pe calculated for each fraction. First, we compared the TLE of the first fraction to assess the intrafraction TLE of the six approaches tested. Second, interfraction TLEs were calculated to evaluate the robustness of the coefficients obtained in the first fraction. The interfraction TLE was defined as the difference between the TLE of a later and the first fraction.

Results: The mean plus two times the standard deviation of the intrafraction TLE was up to 4.3 mm in the CG approaches, while the MRA approaches provided TLEs within 1.5 mm. The mean plus two times the standard deviation of the interfraction TLE did not exceed 1.7 mm in any direction using either approach.

Conclusions: The MRA approach was superior to the CG approach in terms of estimating the target position based on the implanted fiducial markers.

Keywords: Dynamic tumor tracking; Implanted fiducial markers; Multiple regression analysis; Target localization error.