Multimodality image fusion for radiosurgery localisation of large AVMs

Radiosurgery for arteriovenous malformations (AVM) is commonly planned with MR, CT and digital subtraction angiography (DSA) images. T1-weigted MRI and CT provide excellent anatomical information but do not reveal vascular flow. For this reason, DSA images are employed for target localisation, but being 2D projections of 3D structures, they have been shown to be inadequate for the 3D definition of large, complexly shaped AVMs. Magnetic resonance angiography (MRA) does provide vascular volume data, but it suppresses stationary anatomical information and therefore, alone, it cannot be used for treatment planning. To benefit from the functionality of all modalities, the authors have developed a multimodality image fusion environment, which combines 3D MRA and 2D DSA vasculature information with 3D MR/CT anatomical information. Within this environment, MRA/MR/CT datasets are registered and fused by reformatting in common stereotactic space. DSA images are linked to 3D datasets in stereotactic space by recovering their acquisition geometry. The consistency of the recovery procedure is verified by correlation of the DSA images to 2D ray-traced projections of the MRA/MR/CT datasets. Since after the fusion, organ contours drawn on any dataset simultaneously appear on all others, all vascular and anatomical information can be used in the process of target/organ localisation and true 3D delineation. Seven patients that have previously undergone radiosurgery for AVMs are currently being replanned within this environment. The newly defined target volumes are found to differ from those based only on DSA images. This finding can have significant dosimetric implications especially for large AVMs.