Non-rigid full torso respiratory motion correction of SPECT studies

Abstract

Respiratory motion artifact is one of the factors limiting the diagnostic accuracy of SPECT or PET. Full-torso respiratory motion correction in general requires non-rigid respiratory motion estimation and correction due to different organs moving at different rates and possibly directions. In work we presented a method to reduce noise and equalize organ counts in SPECT, to aid non-rigid motion estimation via registration. This is done by transforming the intensities by a function which reduces the noise variance as well as brings the different organ counts to similar levels. We also address the issue of uneven counts in different respiratory states as done in our previous work when addressing rigid registration of the heart for Cardiac SPECT. Then we perform affine registration and B-spline-based free-form deformation on the transformed datasets to obtain the motion estimates. We incorporate the motion estimates within MLEM reconstruction for motion correction as done previously. We applied the method to four actual patient acquisitions. For all four patient datasets we found differences and improvements after corrections. We are also investigating our methodology with NCAT phantom simulations where the truth is know. The NCAT we use was modified to have non-rigid organ motion with respiration.