Triple-Modality Normalized Mutual Information Based Medical Image Registration of Cardiac PET/CT and SPECT Images

An extension of the normalized mutual information similarity measurement was proposed to register cardiac PET/CT and SPECT images representing coronary artery disease in the heart. 23 PET/CT images obtained by a hybrid camera, and their corresponding SPECT were processed. The SPECT was registered to the PET/CT image pairs by a triple NMI based auto-registration. For validation and comparison an triple MI based registration among the PET-SPECT-CT images was performed. Two further dual NMI based auto-registrations between the SPECT-CT and SPECT-PET were performed as well. Based on the comparative results gained by a medical expert, our triple NMI method gave the best transformation parameters with a minimal number of iterations to superimpose the images.