ESTRO 2025 - Abstract Book

S2993

Physics - Image acquisition and processing

ESTRO 2025

1915

Digital Poster Evaluation of biplane 2D digital subtraction angiography and 3D imaging fusion for precise frameless arteriovenous malformation radiosurgery Thierry Gevaert, Marlies Boussaer, Dirk Van den Berge, Selma Ben Mustapha, Mark De Ridder Radiotherapy, Universitair Brussel (UZB), Vrije Universiteit Brussel (VUB), Brussels, Belgium Purpose/Objective: Frameless stereotactic radiosurgery (SRS), enhanced by image-guided positioning systems, is an established treatment method for both benign and malignant brain lesions. For arteriovenous malformations (AVM), digital subtraction angiography (DSA) provides high temporal resolution and valuable dynamic flow information. However, accurately defining the AVM target requires registering orthogonal two-dimensional (2D) DSA views of the nidus with multimodality three-dimensional (3D) imaging. This requirement has traditionally slowed the adoption of frameless techniques, as it typically necessitates a rigid and invasive frame fixed to the skull to immobilize the head. This study aims to evaluate the targeting accuracy achieved by Elements Angio software (Brainlab AG, Munich, Germany), which enables a frameless approach for more streamlined AVM radiosurgery. Material/Methods: A retrospective comparison was performed on ten patients previously treated at our institution using standard frame based SRS. Datasets from digital subtraction angiography in both coronal and sagittal views, along with magnetic resonance angiography (MRA), were imported into the Elements software to assist in targeting the AVM nidus. From the MRA series, a detailed 3D vascular tree was automatically extracted. This model was then co-registered with a selected pair of 2D DSA vascular images through a six-degree-of-freedom rigid registration, using both manual and automated techniques. Target delineation for SRS planning was subsequently performed, and the resulting structure was compared against the reference contour used in the frame-based SRS treatment. To evaluate similarity between the frameless and frame-based target contours, several metrics were applied, including the Dice Similarity Coefficient (DSC), Jaccard Index (JI), Hausdorff Distance (HD), and Mean Distance to Agreement (MDA). Results: No significant difference in AVM nidus volume was found between frame-based and frameless approach (0.79 ± 1.35 cc and 0.83 ± 0.99 cc, respectively). Comparable targeting objects by means of DSC and JI were found: 0.74 ± 0.08 and 0.59 ± 0.10, respectively. Sub-millimetric MDA was found (0.67 ± 0.32 mm). The mean HDshift between individually contoured volumes was 2.97 ± 1.33 mm. These differences were related to the small differences in contouring and segmentation of the objects. Conclusion: The study demonstrates the viability of a frameless approach using Elements Angio for vasculature-based registration, showcasing precise accuracy in target localization for AVM radiosurgery. This method not only confirms its clinical applicability but also provides a significant advantage by eliminating the need for invasive angiography on the day of SRS, streamlining the procedure and enhancing patient comfort.

Keywords: frameless radiosurgery, image registration, AVM