Evaluation of the clinical feasibility of cone-beam computed tomography guided online adaption for simulation-free palliative radiotherapy

Abstract

Background and purpose: Simulation-free radiotherapy, where diagnostic imaging is used for treatment planning, improves accessibility of radiotherapy for eligible palliative patients. Combining this pathway with online adaptive radiotherapy (oART) may improve accuracy of treatment, expanding the number of eligible patients. This study evaluated the adaptive process duration, plan dose volume histogram (DVH) metrics and geometric accuracy of a commercial cone-beam computed tomography (CBCT)-guided oART system for simulation-free, palliative radiotherapy.

Materials and methods: Ten previously treated palliative cases were used to compare system-generated contours against clinician contours in a test environment with Dice Similarity Coefficient (DSC). Twenty simulation-free palliative patients were treated clinically using CBCT-guided oART. Analysis of oART clinical treatment data included; evaluation of the geometric accuracy of system-generated synthetic CT relative to session CBCT anatomy using a Likert scale, comparison of adaptive plan dose distributions to unadapted, using DVH metrics and recording the duration of key steps in the oART workflow.

Results: Auto-generated contours achieved a DSC of higher than 0.85, excluding the stomach which was attributed to CBCT image quality issues. Synthetic CT was locally aligned to CBCT anatomy for approximately 80% of fractions, with the remaining suboptimal yet clinically acceptable. Adaptive plans achieved a median CTV V95% of 99.5%, compared to 95.6% for unadapted. The median overall oART process duration was found to be 13.2 mins, with contour editing being the most time-intensive adaptive step.

Conclusions: The CBCT-guided oART system utilising a simulation-free planning approach was found to be sufficiently accurate for clinical implementation, this may further streamline and improve care for palliative patients.

Keywords: Online adaption; Palliative; Radiotherapy; Simulation-free radiotherapy.

Fig. 1

Brief outline of the CBCT-guided oART workflow and evaluation items included in this study.

Fig. 2

Step-by-step diagram of the kV CBCT-guided oART workflow broken into specific tasks, where the ticks indicate specific timepoints recorded. The duration of the following tasks; influencer generation, influencer edits, other volumes generation and edits, plan generation, and plan review and selection were evaluated in the timing study.

Fig. 3

Comparison of auto-generated (unsupervised) to clinician segmented (supervised) contours using dice similarity coefficient score. The number of fractions evaluated for each contour is reported above the graph. The line denotes the median, and the circle denotes the mean DSC of each contour-type.

Fig. 4

Comparison of CTV and PTV V95% metrics for adapted and scheduled plans of each fraction of oART, evaluated using session synthetic CT and anatomy. The reference plan for each treatment site evaluated on the dCT was also compared. A minimum coverage of CTV V95%>95% represented the primary palliative plan metric. The line denotes the median, and the circle denotes the mean metric performance.

Fig. 5

Comparison of OAR D2cm³ metric for a) single fraction treatment and b) fractionated treatment for adapted and scheduled plans of each fraction of oART, evaluated using session synthetic CT and anatomy. The reference plan for each treatment site evaluated on the dCT was also compared. The line denotes the median, and circle denotes the mean metric performance.

Fig. 6

Duration of each discrete task identified in the kV CBCT-guided adaptive workflow for clinical patient fractions where the line denotes the median, and the circle denotes the mean duration of each step. The mean cumulative time of each task is overlaid.