Three-dimensional image fusion across PET+MRI modalities based on the approach of characteristic coregistration

Because there is no complete three-dimensional (3D) hybrid detector integrated PET+MRI internationally, this study aims to investigate a registration approach for a two-dimensional (2D) hybrid based on characteristic localization to achieve a 3D fusion from the images of PET and MRI as a whole.

A cubic-oriented scheme of "9-point and 3-plane" for a coregistration design was verified to be geometrically practical. Through 3D reconstruction and virtual dissection, human internal feature points were sorted to combine with preselected external feature points for matching process. By following the procedure of feature extraction and image mapping, the processes of "picking points to form planes" and "picking planes for segmentation" were executed. Eventually, the image fusion was implemented at the real-time workstation Mimics based on auto-fuse techniques called "information exchange" and "signal overlay".

A complementary 3D image across PET+MRI modalities, which simultaneously present metabolic activities and anatomic structures, was created with a detectable rate of 56%. This is equivalent to the detectable rate of PET+CT or MRI+CT with no statistically significant difference, and it facilitates a 3D vision that is not yet functional for 2D hybrid imaging.

This cross-modality fusion is doubtless an essential complement for the existing toolkit of a 2D hybrid device. Thus, it would potentially improve the efficiency of diagnosis and therapy for oncology.