Geometric distortion in computed tomography scans for radiotherapy: Quantification and quality assurance implications

Computed tomography (CT) is widely assumed to be geometrically accurate and is commonly used as the reference standard for quantifying geometric distortion in other imaging modalities, particularly magnetic resonance imaging (MRI). However, geometric distortion in CT has not been comprehensively investigated.

The aim of this study was to characterize the performance of a distortion phantom, establish a reference scan and, on this basis, determine and quantify the presence of geometric distortion in CT scans used for radiotherapy (RT) planning.

A CIRS Model 603A Distortion Phantom was employed to quantify geometric distortion. System reproducibility, as well as the effects of phantom setup, scan parameters, fabrication uncertainty, and temporal variation were systematically evaluated. Additionally, the geometric distortion of CT was compared to that of linac-based cone-beam CT (CBCT). Subsequently, geometric distortion magnitudes from eight CT scanners were retrospectively analyzed against both CBCT and a Computer Aided Design (CAD) model.

CBCT demonstrated consistently smaller distortion variation compared with CT. Retrospective analysis revealed that all eight analyzed CT scanners exhibited measurable distortion magnitudes, with mean magnitudes ranging from 0.13 mm to 0.52 mm and maximum distortion magnitudes ranging from 0.45 mm to 1.09 mm when compared to CBCT, or 0.25 mm to 0.82 mm and 0.58 mm to 1.88 mm when compared to the CAD model. Longitudinal couch inaccuracy or non-linear couch sag may constitute the primary contributors to geometric distortion on CT scanners.

Contrary to common assumptions, CT scans used for radiotherapy planning exhibit measurable geometric distortion across clinically relevant fields of view (FoVs). However, current CT QA methodologies and tolerances may lack sufficient sensitivity to detect submillimeter distortion levels relevant to stereotactic radiotherapy (SRT). Recognition and rigorous quantification of CT geometric distortion are essential to inform risk assessment and margin design in high-precision RT.