Clinical Review on Head and Neck CBCT Images

Abstract

Optimal quality assurance is vital in head and neck (HN) radiotherapy (RT), as inferior dosimetry has proven to compromise survival in HN cancer patients. Image guidance using cone-beam computed tomography (CBCT) has enabled precise matching of tumor targets. The objective was to analyze the translational and rotational shifts obtained from HN CBCT in order to develop the 6D HN imaging protocol, henceforth to achieve the best precision radiotherapy. Sixty patients who received treatment from January 2021 to September 2022 were randomly selected - 10 from each sub-site (paranasal sinuses, nasopharynx, oropharynx, oral cavity, larynx-hypopharynx-thyroid and early-stage glottis). Weekly CBCT obtained were analyzed using offline automatic image registration (IR) on an electronic patient information management system. First, 2 regions of interests (ROIs) were defined; one was based on departmental HN IGRT protocol and the other covered the entire planning target volume (PTV). Subsequently, IR was performed with 2 different algorithms; Maximum Mutual Information (MMI) (soft tissue-based) and Chamfer (bone-based). Each CBCT was inspected manually by an individual to avoid inter-observer variability. The shifts were reported as mean ± standard deviation (SD). Paired t-test was used to analyze the variances in translational and rotational displacements between the different ROIs and algorithm. Early-stage glottis was excluded in the ROI comparison as the treatment field was small and the ROI covered the entire PTV. Overall setup errors were 1.2 ± 1.1mm for the translational shifts and 0.7 ± 0.6° for the rotational. The overall frequency of displacement was 95.2% ≤2° and 99.5% ≤5mm. MMI was more constant than Chamfer for HN CBCT matching as reflected by a considerably smaller translational (1.09 ± 1.05mm vs 1.26 ± 1.16mm, p<0.001) and rotational (0.67 ± 0.57° vs 0.77 ± 0.70°, p<0.001) shifts. The comparison of defining ROIs (PTV vs Protocol) was insignificant in the translational shifts (p = 0.057). However, a substantial difference was observed in the rotational displacements (p = 0.003). The analysis was fundamental in identifying the optimal matching tool and it supports the HN 6D imaging protocol. This audit validated the current 5mm tolerance for translational shift and further established a 2° tolerance for the rotational shift for HN 6D CBCT matching. Profoundly, soft tissue localization is of the utmost importance for highly conformal modulated RT, and adaptation of soft tissue matching is strongly encouraged for HN tumors located in these regions.