MR-based Motion Correction Algorithms for PET

MR-based Motion Correction Algorithms for PET

By taking advantage of simultaneity of PET/MR scanners, we have developed several techniques that overcome the limitations of physiological motion to create motion-free PET images. We directly incorporated the motion information generated from MR data into the Expectation Maximization (EM) algorithm used for PET reconstruction. The conventional EM algorithm 

was replaced by:

where X represents the clasical system matrix, f denotes the image to be reconstructed and g contains the acquired emission data. M denotes a motion operator and may include arbitrary motion information. This modification allows the forward and backward projections to be performed along bent lines of response [2], as illustrated in Figure 1.

Because the motion information heavily relies on the respiration signal, the MRI acquisition needs to occur in a continuous, free breathing scheme. We acquired this data in a (golden angel) radial fashion in order to record changes in the k-space center during respiration [1]. Once the respiration signal is extracted, it is used to sort the MRI data into an arbitrary number of respiratory states, usually 5-10. The first respiratory state was identified as the end-expiration phase and used as the reference state. All other states were then registered to this reference state using an algorithm based on optical flow [3]. The spatial transformations estimated during the registration process were then used to construct the motion operator (M). Finally, the MR-derived respiration signal is used to sort and bin the PET data into the matching respiratory states [4] and the extended EM algorithm was performed to reconstruct the motion-compensated PET images.


This is a side-by-side example of reconstruction without motion correction and with:


Principal Investigator: 
Key Personnel: 


Philanthropic Support

We gratefully acknowledge generous support for radiology research at NYU Langone Medical Center from:
• The Big George Foundation
• Raymond and Beverly Sackler
• Bernard and Irene Schwartz

Go to top