ESTRO 2025 - Abstract Book

S1213

Clinical – Lower GI

ESTRO 2025

2547

Digital Poster Early indication of treatment response in rectal cancer using sequential diffusion-weighted imaging at a 1.5 T MR-linac Johanna Austrheim Hundvin 1,2 , Jonas Habrich 3 , Cihan Gani 3,4 , Jörg Assmus 5 , Inger Marie Løes 1 , Sara Pilskog 1 , Kathrine Røe Redalen 6 , Daniela Thorwarth 3,4 1 Department of Oncology and Medical Physics, Haukeland University Hospital, Bergen, Norway. 2 Institute of Physics and Technology, University of Bergen, Bergen, Norway. 3 Department of Radiation Oncology, University Hospital Tübingen, Tübingen, Germany. 4 German Cancer Consortium (DKTK), partner site Tübingen, German Cancer Research Center (DKFZ), Heidelberg, Germany. 5 Centre for Clinical Research, Haukeland University Hospital, Bergen, Norway. 6 Department of Physics, Norwegian University of Science and Technology, Trondheim, Norway Purpose/Objective: Varying tumour response to chemoradiotherapy (CRT) is a challenge in the treatment of locally advanced rectal cancer (LARC) [1, 2]. Early detection of poor responders would enable clinicians to timely adjust the treatment regime by proceeding directly to surgery or escalating the radiation dose [3]. The purpose of this study was to investigate the predictive potential of sequential diffusion-weighted (DW) magnetic resonance imaging (MRI) of LARC during CRT. Material/Methods: The analysis included 15 patients with LARC (ten male, five female) prospectively enrolled for treatment on a 1.5 T MR-Linac between October 2018 and June 2021. All patients were prescribed long-course CRT to the pelvis with 50.4 Gy in 28 fractions. Following surgery, the tumour response was classified as good (Dworak 3 or 4) or poor (Dworak 1 or 2) based on pathologic tumour regression [4]. DW images (DWIs) were acquired on a diagnostic 3 T MRI scanner (Siemens Healthineers) for baseline and the 1.5 T Unity MR-Linac (Elekta, AB) over the course of MR-guided radiotherapy, with b-values=0, 150/200, 500 and 800 s/mm 2 . The primary tumour was segmented axially on the high b-value DWI with the corresponding T2-weigthed image as reference. An in-house python script calculated the apparent diffusion coefficient (ADC) based on a mono-exponential model using b-values=150/200 and 500 s/mm 2 . Development of linear mixed-effects modelling was applied to analyse changes in mean tumour ADC as a function of fraction and outcome. Model improvement was assessed through ANOVA analysis and level of significance set to p<0.05. Results: Six patients had a good response, whereas nine patients had a poor response. There was a significant difference in ADC at baseline between the two groups (mean Responders = 1.3×10 -3 mm 2 /s, mean Non-responders = 1.1×10 -3 mm 2 /s, p<0.05; Figure 1A). The linear mixed-effects model revealed a significant impact of the interaction term, implying a steeper slope in ADC change among responders than non-responders (β=0.01, p<0.01). Investigation of the number of fractions included in the model did not result in a clear optimum but indicated that a treatment response could be determined already after the five first fractions (Figure 1B and Figure 2).