Adaptive Emergency Radiotherapy Using Diagnostic Computed Tomography Imaging: Initial Results from a Retrospective Geometric Evaluation and Clinical Considerations

Background: Radiotherapy commonly requires a computed tomography (CT) dataset for treatment planning. Artificial intelligence-supported radiotherapy systems now enable online adaptive workflows that reduce uncertainties resulting from the use of diagnostic instead of planning CT images. The aim was to identify patient groups that may particularly benefit from adaptive emergency treatments. Methods: In this retrospective exploratory study, 16 patients comprising 21 planning target volumes (PTVs) were evaluated. PTVs were classified as predominantly soft-tissue or bony. To verify the geometric agreement between scheduled and adapted treatment plans, these subgroups were analyzed using absolute planning target volume change, the Dice similarity coefficient (DSC), and the 95th percentile Hausdorff distance, and their association with V95% PTV coverage was evaluated. Results: Bony targets showed higher geometric reproducibility than soft-tissue-dominant lesions, with higher median DSC values (0.96 vs. 0.88) and lower median HD95 values (2.6 mm vs. 5.5 mm). Online adaptation improved coverage in 20 of 21 PTVs, while predominantly soft-tissue PTVs showed potentially greater benefit from online adaptation. Conclusions: Soft-tissue-dominant PTVs showed greater geometric variability and potentially greater benefit from adaptation, whereas selected bone-dominant cases may be suitable for faster conventional workflows. Given the small exploratory cohort, these results require further confirmation.