A Multimodal Image Registration and Fusion Methodology Applied to Drug Discovery Research

The development of novel methodologies that utilize various non-invasive imaging modalities has resulted in their increased use and relevance in the pie-clinical phase of pharmaceutical compound development. Scientific questions that may benefit from the use of imaging techniques are often more robustly addressed with data acquired from complementary imaging modalities, intensifying the need for performing cross-modality image registration and fusion. In this work, a methodology for multimodal coregistration and fusion of MRI and PET volumes is presented with the aim of visualizing the biodistribution of a radiolabeled compound (from PET) in a high-resolution anatomical reference image (from MRI). The use of an animal platform and fiducial markers for defining spatial correspondences is considered first, followed by a preliminary automated alignment approach that calculates and matches shape-based features using Parzen estimators and genetic algorithm optimization. Experimental results are also presented, followed by concluding remarks and future perspectives on our work.