Multimodality image fusion of the liver using structure-guided deformable image registration in velocity AI—what is the preferred approach?

Abstract

This study looked at the impact of using different volume segmentation in deformable image registration (DIR) for multimodality imaging, specifically planning computed tomography (CT) and post stereotactic body radiation therapy (SBRT) magnetic resonance imaging (MRI), where liver was the target of the registration. Planning CT and post-SBRT MRI scans for 9 previously treated patients were used in this study. The MRI scan was deformed on to the planning CT using structure-guided DIR using the commercial software Velocity AI. Three different deformation methods were employed based on different contoured regions of the liver and liver itself. The Dice similarity coefficient (DSC) was quantified for all of the structures within the liver along with the average voxel displacement. The registration method that was based only on the liver contours had the largest DSC for some of the internal liver structures, which suggests a preferred approach when using structure-guided DIR for the liver.