ESTRO 2024 - Abstract Book

S3818

Physics - Image acquisition and processing

ESTRO 2024

1174

Digital Poster

Assessment of residual MRI geometric distortions on brain and spine stereotactic treatments

Karima Tamsamani 1,2 , Aurélie Tournier 1 , Thomas Puiseux 2 , Soleakhena Ken 1,3 , Ramiro Moreno 2 , Laure Vieillevigne 1,3

1 Oncopole Claudius Regaud, Medical Physics, Toulouse, France. 2 Spin Up ALARA, R&D, Entzheim, France. 3 Centre de Recherches de Cancérologie de Toulouse, INSERM, Toulouse, France

Purpose/Objective:

Stereotactic radiotherapy is an external radiotherapy technique that involves high doses to target volumes, while minimizing the dose received by organs at risk (OAR). CT images remain the reference images for dose calculation in radiotherapy but MRI images are often necessary to define the target volumes as they offer better spatial resolution and soft tissue contrast. However, even after vendor distortion corrections, residual distortions might still be present and could alter the representation of anatomical images [1]. Thus, these distortions can impact the contouring of target volumes and consequently the treatment plans, in particular for small targets. The aim of this study was to quantify residual gradient nonlinearity distortions for stereotactic treatments for brain and spine tumors and to evaluate their impact on dose distributions.

Material/Methods:

The Cartesian3D body phantom (Spin Up, ALARA, France) measuring 45 cm x 39 cm x 37 cm was used to generate the corrected gradient non-linearities 3D distortion map. This phantom was made of stacked polystyrene panels and contained around 2 000 beads visible on CT and MRI. The 3D distortion map as a function of the position of each bead center was obtained by comparing the reference CT to the MRI images acquired for the clinical sequences (T1 3D Gradient Echo for brain and T1/T2 3D SPACE for spine). Twenty patients with multiple brain tumors preferably located in the periphery of the brain and five patients with spine tumors were selected. Their MRI-contoured DICOM RTStructs were corrected by applying phantom-based 3D distortion maps through the Spin Up CartesianRT software. Differences between structures (target volumes and OAR close to target) before and after the correction of the distortion were evaluated with the Dice score and the Hausdorff distance. The impact on the dose distribution and on the dose to 98% of the target volume were assessed.

Results:

The worst case, for brain, gave a Dice index and a Hausdorff distance close to 0.7 and to 1.3 mm, respectively when the distance from the magnet isocenter was 90 mm. A 1 mm limit was achieved on average at a distance of 70 mm from the center of the magnet. For the 20 patients, the D98% of the target volume was on average reduced by 1% with a maximal deviation of 2.5%.

For the spine tumors, distortion mainly affected the spinal cord due to its position at the edge of the field of view and its geometric shape. The Hausdorff distance was found up to 4 mm at around 60 mm from the magnet isocenter. For