ESTRO 2024 - Abstract Book

S3921

Physics - Image acquisition and processing

ESTRO 2024

Radiotherapy (RT) trends today tend to go towards higher doses and smaller margins. This demands high accuracy throughout the treatment chain with specific knowledge about related uncertainties. One potential uncertainty comes from the inter-modality image registration between magnetic resonance (MR) and computed tomography (CT) images, routinely performed for brain cancer patients in the conventional RT workflow. The elimination of this registration has been one of the arguments for MRI-only RT, but is it really motivated?

The aim of this work was to investigate the uncertainty related to the image registration between MR and CT images of the brain using a clinical registration method.

Material/Methods:

Image registrations according to clinical routine were performed by an experienced observer. Forty-five paired MR CT image sets of patients with intracranial tumours were evaluated (glioma (n=23), metastases (n=19), meningioma (n=2), astrocytoma (n=1)). For each patient, a manual match box was inserted to include the brain and an automatic rigid, mutual information-based, registration was performed in Eclipse (v 15.6, Varian Medical Systems, Palo Alto, CA, US). The observer registrations were compared to those performed in clinical practice by calculating the difference in six dimensions (translations and rotations) between the registrations. Subsequently, the centre of mass (CoM) displacement, DICE-index and Hausdorff distance of the gross tumour volume (GTV) and the planning target volume (PTV) were calculated for patients with difference vectors larger or equal to 1 mm. Patients with missing structures were not included.

Results:

The differences between observer and clinical registrations are presented in Figure 1. All rotational differences were within ±1 degree and all translational differences, except for the superior-inferior (S-I), were within ±1 mm. In the S-I direction, absolute differences up to 2.1 mm were found.