ESTRO 2025 - Abstract Book

S61

Invited Speaker

ESTRO 2025

4743

Speaker Abstracts Towards biological image guided adaptive radiotherapy using quantitative MRI in clinical multicentre trials Petra J van Houdt Radiation Oncology, the Netherlands Cancer Institute, Amsterdam, Netherlands Abstract: Quantitative imaging biomarkers (QIBs) derived from MRI techniques, like diffusion-weighted imaging and dynamic contrast-enhanced MRI, have the potential to be used for personalized treatment of cancer patients. An exciting prospect is the use of QIBs to adapt the radiation treatment in a strategy called biological image guided adaptive radiotherapy (BIGART). This might involve dose escalation or de-escalation by modifying the number of treatment fractions and/or total dose to the tumor or even modulate the dose distribution using the spatial information in QIB maps [1]. To implement BIGART, repeated MRI measurements are needed to measure the changes in biological characteristics. In principle this can be done on regular diagnostic MRI systems (or MR simulators). However, the introduction of MRI-guided RT systems has increased the potential to bring BIGART into clinical practice, as they have the capability to adapt treatments during each fraction. Clinical validation of QIBs for BIGART has started with establishing changes in quantitative MRI (qMRI) values during the course of radiotherapy. The first studies have identified such changes in rectal cancer, brain cancer, and soft tissue sarcoma in small mostly single-center patient cohorts [2]. The next step would be to validate the prognostic and predictive value of these promising qMRI techniques in larger multicenter trials. However, performing qMRI in a multicenter setting is challenging. Variations in qMRI parameter values can be attributed to acquisition differences related to variations in MR hardware from different vendors, vendor-specific implementations of a qMRI technique, as well as choices in protocol settings. Also differences in analysis approaches to calculate qMRI maps from the acquired data attributes to variations often seen in literature. Those involve the choices that are being made for the data analysis, the software that is being used, algorithm implementations, and inter-subject variation in delineation of the region of interest. Therefore, technical validation is needed to make sure that the qMRI measurements can be done anywhere and give comparable results [3]. Technical validation aims to measure the bias, repeatability, and reproducibility of the qMRI values. Bias describes the difference of a measurement from the ground truth and is typically established with dedicated qMRI phantoms. Repeatability relates to repeated measurements over time on the same subject and machine and is especially important to distinguish a treatment-related change from random day-to-day variations in qMRI values. Repeatability is typically assessed with in-vivo test-retest studies. Reproducibility is related to measurements under different conditions, for example, in a multicenter setting with different MR systems. Both repeatability and reproducibility can be measured with phantoms, but in-vivo measurements will give a more realistic estimate of the precision in a clinical trial setting. How good the technical performance of the qMRI measurements needs to be depends on the expected effect size that we want to be able to measure. This might differ per application and is usually not known up front. There are different ways to minimize the variation of qMRI values in a multicenter setting. First, a dedicated QA procedure should be in place before a trial starts to measure and potentially minimize differences in qMRI parameters between centers. Guidelines and sharing of qMRI protocol settings will help to minimize these differences. These QA measurements will give insight whether scaling of the in-vivo qMRI values might be needed on institutional basis. Furthermore, the use of the same, preferably open source, pipelines for data analysis will reduce any variation introduced by analysis choices. In conclusion, qMRI has the potential to be used as biomarkers for BIGART, especially with the introduction of MRI guided RT systems this is within reach. Clinical validation is currently ongoing but mostly in small single-center cohorts. To make the next steps, multicenter validation is needed. To improve the success of multicenter trials, technical validation should be integrated in trials as well as harmonization of acquisition and analysis protocols. References [1] van Houdt PJ, Saeed H, Thorwarth D, et al. Eur J Cancer. 2021;153:64-71

[2] van Houdt PJ, Li S, Yang Y, van der Heide UA. Semin Radiat Oncol. 2024 Jan;34(1):107-119. [3] O’Connor JPB, Aboagye EO, Adams JE, et al. Nat Rev Clin Oncol. 2017;14(3):169-186.