Impact of eFOV CT distortion on surface-guided radiotherapy registration: Phantom-based analysis

This study evaluated the impact of extended field-of-view (eFOV) CT distortion on the alignment accuracy of surface-guided radiation therapy (SGRT) using CT-based reference surfaces, and examined the influence of beam center location.

An anthropomorphic phantom, segmented into chest, abdomen, and pelvis, was scanned at the center position (sFOV) and at four off-center positions. The center corresponds to the CT bore isocenter, while the off-center positions were obtained by placing the phantom boundary 315 mm or 390 mm to the left or right of the CT isocenter to simulate eFOV conditions. Geometric distortion was assessed by measuring external body contour volume. Treatment plans were generated with beam centers at the isocenter and at off-center positions of 8 cm left and right. SGRT registration accuracy was evaluated from SGRT alignment error and cone-beam CT (CBCT) alignment error differences using axis-specific mean absolute errors (MAEs) and overall translational and rotational root mean square errors (RMSEs).

Distortion was most pronounced in the chest when the phantom boundary was 390 mm right of the isocenter, with a 12.55% volume reduction. In contrast, abdomen and pelvis volume changes were within 1% across all positions. At the isocenter, MAEs increased at 390 mm off-center positions, with chest showing lateral/vertical errors and abdomen exhibiting longitudinal/rotational deviations. When the beam center was 8 cm lateral and adjacent to a distorted region, errors increased, while contralateral placement showed errors comparable to the isocenter. Maximum RMSEs of 10.98 mm (translation) and 5.88° (rotation) occurred in the abdomen when beam centers overlapped distorted regions, exceeding typical clinical thresholds (5 mm, 2°). The pelvis showed relatively stable alignment with RMSEs within 2.5 mm and 1.5°.

This phantom study demonstrated that eFOV reconstruction can compromise SGRT alignment accuracy, particularly when beam centers are located near distorted regions. Alignment remained within clinically acceptable limits when reference points were sufficiently distant from distortion. These findings provide important baseline data under controlled conditions, and clinical validation with patient cohorts is warranted to confirm their generalizability.