ESTRO 2024 - Abstract Book

S4146

Physics - Intra-fraction motion management and real-time adaptive radiotherapy

ESTRO 2024

Georgios Tsekas, Cornel Zachiu, Gijsbert H Bol, Madelon van den Dobbelsteen, Lieke T C Meijers, Astrid L H M W van Lier, Hans C J de Boer, Bas W Raaymakers

UMC Utrecht, Radiotherapy, Utrecht, Netherlands

Purpose/Objective:

Intrafraction motion poses an important challenge in radiotherapy of prostate cancer patients and leads to treatment uncertainties. Recently, a 2D cine MRI-based motion monitoring package was developed that enables real-time motion tracking and intrafraction drift corrections [1, 2]. Compared to the current clinical workflow in our department [3], real-time motion monitoring allows for faster intrafraction adaptations (less than 1 minute) and can offer a more patient-tailored solution, as prostate motion can largely differ from fraction to fraction. To that end, the purpose of this work was to estimate the accuracy of 2D cine MRI-based intrafraction motion monitoring in the context of adaptive radiotherapy of prostate cancer patients on the 1.5T MR-linac using an independent registration procedure.

Material/Methods:

In our study we used the recently released Comprehensive Motion Monitoring (CMM) software (Elekta AB, Sweden), currently clinically available on the 1.5T MR-linac. CMM performs real-time organ tracking using a set of interleaved orthogonal 2D cine MRI images (coronal and sagittal planes) and can be used for respiratory gating. In addition, it enables intrafraction drift corrections in the presence of large motion events. A drift correction results in stopping the treatment and applying a rigid translation to the remainder of the treatment plan prior to continuing the tumour irradiation. In order to evaluate the 2D tracking precision of the CMM software we initially benchmarked it against elastix, an independent, well established image registration algorithm [4]. The two algorithms were used for tracking the intrafraction motion of the prostate using 2D cine MRI scans of prostate cancer patients treated in our clinic. The scan acquisition took place during the plan optimization phase of a treatment session and the CMM-default balanced turbo field echo (bTFE) MRI sequence was used. A total of twelve fractions corresponding to eight prostate cancer patients with an average duration of 8.5 minutes were used for our analysis. We initially compared the resulting motion signals of the two algorithms in terms of positional mismatch per motion direction. The mismatch was defined as the difference between the elastix- and CMM-based tracked motion for each image frame. Furthermore, a set of virtual drift corrections were generated from the motion signal of each algorithm: Initially, the first sample was considered as the reference position of the prostate. A virtual drift correction was triggered when the prostate centroid drifted away from its reference position for more than a predefined distance threshold and for at least 60 seconds. The reference position was reset after a drift correction to the last position before the plan adaptation. The decision thresholds were set to 1 mm and 2 mm respectively for our experiments as these values are typically used in a clinical setting. The total 3D displacement of the prostate centroid was used as a distance criterion for deciding upon the need of drift corrections as defined below:

2 + (AP

2 + (SI

2 )

displacement = √( (LR tracked - LR ref )

tracked - AP ref )

tracked - SI ref )

, where the differences of the tracked and reference positions of the centroid for the Anterior-Posterior (AP), Left Right (LR) and Superior-Inferior (SI) directions were used.