Vienna, Austria

ESTRO 2023

Session Item

Sunday
May 14
10:30 - 11:30
Strauss 3
Advancements in RTT practice
Loes Bulthuis, The Netherlands;
Michelle Leech, Ireland
Proffered Papers
RTT
11:20 - 11:30
Verification of rectal cancer target auto-contouring by trained RTTs in online adaptive MRIgRT
Anja Betgen, The Netherlands
OC-0466

Abstract

Verification of rectal cancer target auto-contouring by trained RTTs in online adaptive MRIgRT
Authors:

Anja Betgen1, Lisa Wiersema1, Nicole Ferreira Silvério1, Rita Simões1, Corrie Marijnen1, Femke Peters1, Uulke, A van der Heide1, Tomas Janssen1

1The Netherlands Cancer Institute, Department of Radiation Oncology, Amsterdam, The Netherlands

Show Affiliations
Purpose or Objective

For rectal cancer patients, anatomical changes can be significant and daily plan adaptation allows for a reduction of treatment margins. Adjustments of the target volume are normally performed by Radiation Oncologists (RO), requiring their presence during each fraction. Auto-contouring can help streamline the process, but for target volumes such solutions are not widely available. To reduce the burden for the RO nonetheless, we trained RTTs to take over this task. The contours generated by RTTs during online adaptation were used to develop an auto-contouring solution for the mesorectum.  Since the auto-contour still needs to be verified before clinical use, it does not remove the need for the RO to be present. The aim of this study was to show that the combination of auto-contouring with a verification by trained RTTs, leads to an efficient online adaptive workflow.

Material and Methods

The training program consisted of education by a radiologist, delineation instructions plus a demonstration by an RO and written instructions. After training, ten patients were randomly selected for re-delineation by an RTT, using the deformed target volume as prior. These re-delineations were checked by an RO and after approval RTTs were allowed to perform the re-delineation online, followed by an offline check before the next fraction. An in-house developed auto-contouring algorithm was trained for the mesorectum based on the on-line contours. Trained RTTs verified and/or corrected the auto-contour in an offline setting and scored the accuracy of the auto-contouring as well as the timing for verification and/or correction. To evaluate the potential time gain of this workflow, the timing of re-delineation was also scored during online adaptation. The timing for manual re-delineation was performed for 13 patients (51 fr), validation of the auto contour was performed for 18 patients (50 fr).



Results

Four RTTs were trained to perform online re-delineation. In total 45 patients (190 fr) were re-delineated by RTTs, all delineations were considered clinically acceptable. Online re-delineation of the target volume took on average 7:34 min. Minimal time was 2:40 min, maximum time 14:58 min. For offline validation of the auto-contour these numbers were resp. 2:45 min, 0:35 min and 8:35 min. Only for 2 fractions the auto-contour was judged unacceptable. No corrections were required for 9 fractions; while the remaining fractions only required minimal correction (29) or a mix of large and small corrections (9).

Conclusion

RTTs are skilled to delineate the target volume in rectal cancer patients during the MRI-guided adaptive process in absence of an RO. This allows for a reduction in logistic complexity and cost, making online adaptive radiotherapy available for more patients.  Auto-contouring  of target volumes can assist RTTs in this process to reduce delineation time. Trained RTTs remain important to guarantee correct contours as a start for online plan adaptation.