Coregistration of PET/MR brain images by multi-feature correlation matching

Medical images analysis is becoming increasingly important in clinical applications. One of the active research areas in medical image analysis is image coregistration which involves information fusion of tomographic diagnostic images obtained from different modalities. The authors present a novel MRI and PET brain image coregistration technique using binary correlation matching based on multiple image features. A two-feature image is formed by extracting edge and region information from PET and MR images. By unifying the pixel intensities and anatomical information in the PET and MR images, the multi-feature PET and MR images are then cross-correlated to find the minimum mismatch energy which corresponds to best matching transformation. The consistent nature of the misregistration curves and comparative studies show that the authors´ matching technique results in a robust and accurate coregistration of MRI and PET images