ESTRO 2024 - Abstract Book

S3986

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2024

In fourteen pediatric cancer patients receiving radiotherapy, we found a significant reduction of the spine length (T11– L4) between the CBCTs acquired in the morning and in the afternoon. No correlation between the difference in acquisition time and spine length variation was found, but our results implicate that diurnal spine length reduction should be taken into account by preferably planning the treatment fractions at the same time of the day as the pCT was acquired. Hence, diurnal spine length reduction is a factor to take into account when treating with multiple fractions a day, or when fractions are planned at different time slots during the course of treatment.

Keywords: Pediatric radiotherapy, IGRT, spine length

References:

1. Van Deursen, L. L., Van Deursen, D. L., Snijders, C. J. & Wilke, H. J. Relationship between everyday activities and spinal shrinkage. Clin. Biomech. 20, 547–550 (2005).

594

Digital Poster

An automated assessment pipeline to identify prostate treatments that need adaptive radiotherapy.

Emily Russell 1 , Christopher O'Hara 1 , Sebastian Andersson 2 , Ann Henry 1,3 , Richard Speight 1 , Bashar Al-Qaisieh 1 , David Bird 1 1 Leeds Cancer Centre, Medical Physics, Leeds, United Kingdom. 2 RaySearch Laboratories, Research Department, Stockholm, Sweden. 3 University of Leeds, School of Medicine, Leeds, United Kingdom

Purpose/Objective:

When anatomical change is detected during on-treatment imaging there are uncertainties in deciding whether adaptive radiotherapy (ART) is required due to the subjectivity of visual assessment. ART decision-making also requires substantial time and resources from a multidisciplinary team. This project aimed to develop and validate an automated pipeline for 60 Gy in 20 fraction prostate cancer treatments which could accurately determine which patients could benefit from ART using synthetic CTs (sCTs) generated from on-treatment Cone-Beam CT (CBCT) images. A secondary aim was to quantify the potential time saving of using this automated pipeline vs. manual and current clinical methods.

Material/Methods:

A pipeline was developed utilising Python code to convert CBCTs to dosimetrically accurate sCTs utilising a Cycle GAN inspired deep-learning model within the RayStation treatment planning system (TPS). Deformable registration was used to map contours from the planning CT to the sCT, and the treatment plan was recalculated on each sCT. A traffic light model was created corresponding to relevant clinical goals, where an sCT was given a ‘red’ outcome when a