ESTRO 2024 - Abstract Book

S3987

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2024

mandatory bowel or CTV dose objective failed (if it was met in the original plan). This would suggest ART may be required. An sCT was given a ‘green’ outcome if no objectives failed, suggesting no significant anatomical changes had occurred. Figure 1 describes the automated steps taken by the pipeline. The pipeline was applied to 31 patients (6 which had ART) for every acquired CBCT (resulting in 230 sCTs total), to assess the pipeline accuracy, comparing the clinical outcome (ART or no ART) to the pipeline outcome. Pipeline accuracy was assessed using a receiver operating characteristic (ROC) curve. Sensitivity/specificity analysis identified an optimal threshold of ‘red’ sCTs to indicate ART was required. A timing assessment compared the automated pipeline to a manual pipeline pathway and the current clinical ART assessment pathway to quantify time saving. Pipeline contour accuracy was investigated through visual assessment of contours by 2 physicists using a Likert scale. The impact of erroneous rectum contours was investigated by manually correcting contours and re-running the pipeline to assess whether errors changed the pipeline outcome.

Results:

The pipeline was capable of distinguishing between patients who required ART from those who did not; 74.4% of sCTs from ART patients received a red traffic light result, compared to 6.4% of sCTs from non-ART patients, which is statistically significant (p<0.05). The ROC area under curve was 0.98, demonstrating the pipeline’s high performance at distinguishing between ART and non-ART patients. Sensitivity/specificity analysis indicated a threshold of 2 red sCTs would be optimal for identifying when ART was required. Figure 2 shows the pipeline outcome data for all patients that received a red sCT and the treatment fraction at which red sCTs occurred. It demonstrates the clinical impact of the pipeline, where the majority of ART patients had high numbers of red sCTs. With a threshold of 2 red sCTs applied, 5/6 ART patients would have been identified earlier using the pipeline. One ART patient would not have been identified; however this is because they were replanned at the same fraction as their first red sCT. Three non-ART