Impact of CBCT Reconstruction and DIR Algorithm on Supraglottic Contour Accuracy In Oropharyngeal Cancer
Abstract
Purpose
Accurate delineation of anatomy on cone-beam CT (CBCT) is essential for image-guided and adaptive radiotherapy in head and neck cancer, yet image quality and deformable image registration (DIR) performance remain limiting. This study evaluated accuracy of supraglottic contours propagated from planning CT (pCT) to CBCT using two commercial DIR-based auto-contouring algorithms and assessed impact of CBCT reconstruction method on contour accuracy.
Methods
Sixteen patients with oropharyngeal cancer underwent intensity modulated radiation therapy and daily CBCT images were retrospectively analyzed. Supraglottic contours delineated on pCT were propagated to first-fraction CBCT using two DIR-based auto-contouring algorithms: a free-form, intensity-based algorithm (FF-IBA, MIM ver7.4.3) and a multi-resolution elastic B-spline algorithm with mutual information (MR-EBSA, Velocity ver4.1). Contours were compared against reference CBCT contours delineated by an expert radiation oncologist. Eight CBCTs were reconstructed using standard CBCT (sCBCT) and eight using iterative CBCT (iCBCT). Accuracy was evaluated using Dice similarity coefficient (DSC), average surface distance (ASD), 95th percentile Hausdorff distance (HD95), center-of-mass displacement (CM), relative/absolute volume difference (RVD/AVD).
Results
Across all patients, FF-IBA demonstrated higher contour agreement than MR-EBSA, with mean DSC of 0.65 ± 0.10 versus 0.59 ± 0.12 and lower ASD (0.74 ± 0.40 mm vs 0.90 ± 0.71 mm). iCBCT reconstruction improved contour accuracy for both algorithms, yielding higher median DSC values and reduced ASD and HD95 compared with sCBCT. Median DSC increased from approximately 0.55 to 0.69 for MR-EBSA and from 0.63 to 0.71 for FF-IBA. CM remained comparable between reconstruction methods (median <3 mm), indicating no systematic geometric bias. Volume agreement improved with iCBCT, with reduced variability, fewer extreme RVD outliers and smaller AVD values.
Conclusion
Supraglottic contour accuracy of deformable propagation from pCT to CBCT is influenced by DIR algorithm and CBCT reconstruction method. iCBCT reconstruction improves boundary agreement and volume consistency without introducing systematic geometric bias.