Preliminary evidence of systematic longitudinal offsets in couch corrections associated with CT localizer scan-line placement in head-and-neck image-guided radiotherapy: a brief research report

Background Computed tomography (CT) simulation transfers external setup references into treatment planning and image-guided radiotherapy, but the effect of localizer scan-line placement relative to radiopaque markers remains uncertain. Objective This study examined associations between scan-line placement and directional differences in longitudinal couch corrections during head-and-neck image-guided radiotherapy. Methods This single-center study included 104 patients and 3,115 pre-treatment cone-beam CT records. Two operational cohorts used cranial-edge (A) or caudal-edge (B) scan-line alignment. A retrospective routine center-aligned cohort provided a clinical reference (C). The primary endpoint was patient-level treatment-course mean longitudinal couch correction (LNG), retaining treatment-system direction. Absolute LNG, events exceeding 5 mm, mixed-effects models, and an exploratory marker-visibility centroid were analyzed. Results Mean LNG corrections were −0.312, 0.625, and 0.327 mm in groups A, B, and C. The patient-level A-B difference was −0.937 mm [Bonferroni simultaneous 95% confidence interval (CI), −1.865 to −0.009; Holm-adjusted P = 0.0472]. The mixed-effects estimate was −0.930 mm (Bonferroni simultaneous 95% CI, −1.887 to 0.027; Holm-adjusted P = 0.0595). No group difference in mean absolute LNG was detected ( P = 0.5068). Each 1-mm increase in the marker-visibility centroid was associated with a 0.519-mm increase in LNG (HC3 95% CI, 0.367 to 0.671). Conclusion The evaluated workflows showed a small directional difference in pre-treatment longitudinal couch corrections, with statistical support varying across analyses. No difference in mean absolute correction was detected. These hypothesis-generating findings require further validation because scanner–operator confounding, unverified geometry, observer bias, and consent-related selection limit causal interpretation.