Adaptive radiotherapy triggering for nasopharyngeal cancer based on bayesian decision model

Abstract Objective. To develop a Bayesian decision model for adaptive radiotherapy (ART) in nasopharyngeal cancer (NPC) that balances clinical capacity of ART and inter-fraction dosimetric changes. Approach. A retrospective analysis was conducted on 84 fractions from 17 NPC patients treated with intensity-modulated radiotherapy using a CT-Linac. Fourteen patients were included for the model construction, and three for validation. Daily diagnostic-level CT images were rigidly registered to the planning CT for regions of interest and treatment plan propagation. The propagated contours were reviewed and refined by radiation oncologists. For each daily CT, percentage differences in 27 dose metrics were compared to the original plan. Composite scores of dose differences were developed using factor analysis on planning target volume (PTV) and organ at risk (OAR) dose metrics. These scores were integrated into a Bayesian decision model, which incorporated a subjective trigger rate to determine the initiation of ART. Main results. The model generated individualized re-plan strategies based on composite scores for PTV or OAR, with trigger rates ranging from 10% to 60%. In the validation with 14 fractions, significant anatomical and dosimetric variations were identified. At a 30% trigger rate, only one fraction was misclassified. Significance. It is feasible to employ a Bayesian decision model for ART, merging subjective clinical insights with objective dosimetric data to refine re-planning decisions.