Evaluation of dose accumulation methods and workflows utilising cone beam computed tomography images

Introduction Various adaptive radiation therapy (ART) methods have emerged, with little consensus amongst the literature as to which is most appropriate. This study aimed to compare dose mapping (DM) versus Monte Carlo recalculation (MCR), using cone beam computed tomography (CBCT) images when utilised in automated ART dose accumulation workflows in the MIM Maestro software package. Methods The treatment plans for 38 cancer patients (19 prostate and 19 head and neck cases) were used to perform DM or MCR retrospectively upon CBCTs acquired during treatment, which were then deformably registered to the planning CT (DR‐pCT) to facilitate dose accumulation. Dose–volume and region‐of‐interest data were extracted for the planning target volumes and organs at risk. Intraclass correlation (ICC) values and Bland–Altman plots were utilised to compare DM versus MCR doses on the CBCT images as well as CBCT versus DR‐pCT doses. Results When comparing DM and MCR on CBCTs, the differences across dose level mean dose differences were mostly within a ±5% level of agreement based on the Bland–Altman plots, with over 67% of ICC values over 0.9 and indicative of good correlation. When these distributions were deformed back to the planning CT, the agreement was reduced considerably, with larger differences (exceeding ±5%) resulting from workflow‐related issues. Conclusion The results emphasise the need to consider and make adaptations to minimise the effect of workflows on algorithm performance. Manual user intervention, refined departmental protocols and further developments to the MIM Maestro software will enhance the use of this tool. Various adaptive radiation therapy (ART) methods have emerged, with little consensus amongst the literature as to which is most appropriate for clinical use. This study compared dose mapping to Monte Carlo recalculation on a cohort of 38 prostate and head and neck patients using cone‐beam computed tomography images. The results emphasise the need to consider and make adaptations to minimise the effect of workflows on algorithm performance.