ESTRO 2025 - Abstract Book

S3000

Physics - Image acquisition and processing

ESTRO 2025

2029

Digital Poster Preliminary Results from Consensus Measurement Methods for Longitudinal relaxometry on the 1.5T Unity MR-linac: a multi-institution comparison Michael J. van Rijssel 1 , Travis C. Salzillo 2 , Ergys D. Subashi 2 , Erik van der Bijl 3 , Warren Foltz 4 , Marijn Kruiskamp 5 , Astrid L.H.M.W. van Lier 1 , Hanna Liu 1 , Eric Paulson 6 , Bjorn Stemkens 7 , Mariëlle E.P. Philippens 1 1 Radiotherapy, UMC Utrecht, Utrecht, Netherlands. 2 Radiation Oncology, UT MD Anderson Cancer Center, Houston, USA. 3 Radiotherapy, RadboudUMC, Nijmegen, Netherlands. 4 Radiation Oncology, UHN, Toronto, Canada. 5 MR Therapy, Philips, Best, Netherlands. 6 Radiation Oncology, MCW, Milwaukee, USA. 7 MR/RT, Elekta, Best, Netherlands Purpose/Objective: The Elekta Unity MR-Linac combines a 1.5T MRI with a linear accelerator for simultaneous MR-imaging and external beam radiotherapy [1]. It provides the potential for online adaptive radiotherapy planning and delivery based on quantitative imaging biomarkers (QIBs) [2]. A multi-institutional task group now pursues a consensus protocol for T1 and T2 quantification in a fast but reliable manner, enabling relaxation based QIBs. This work summarizes the progress towards a T1 consensus protocol, including identification of candidate techniques and their evaluation using the ISMRM-NIST phantom [3], which contains a series of reference samples for T1 and T2. Material/Methods: The task group developed a protocol for T1 mapping after several consensus meetings and is evaluating inter- and intra institutional reproducibility and accuracy of 4 methods (Table 1).The ISMRM-NIST phantom was imaged thrice in institute A on two MR-Linacs and twice in institute B. Gold-standard methods, IR and MIX, were performed in institute A only, with IR acquisition limited to the T1 reference spheres.

Image processing utilized established techniques for T1 mapping, including an RF flip angle correction for incomplete RF spoiling at short TR for DESPOT1 [8]. Analysis was limited to the central slice, consisting of a one pixel erosion from manually segmented ROIs drawn on the individual reference spheres. For each sphere, bias was calculated as the difference between mean T1 and the vendor-supplied reference. Repeatability (within-subject deviation) and reproducibility (within-subject coefficient of variation) were calculated according to QIBA recommendations [9] for multi-institutional data (only for DESPOT1 and MIRACLE).