Motion-compensated temporal summation of cardiac gated SPECT images using a deformable mesh model.

We propose a motion-compensated non-rigid summation method for noise reduction in cardiac gated SPECT. This approach generates a static SPECT image containing counts from all frames of the gated sequence while accounting for heart motion to avoid motion-blur artifact. Static cardiac images typically suffer from heart motion occurring during acquisition which introduces the so-called motion blur artifact. Gated acquisitions, on the other hand, are characterized by lower counts in each individual frame, thus resulting in noisy images. Methods have been proposed to sum the gated sequence along the time dimension while accounting for heart motion, but they do not account for partial volume effect, manifested by an intensity increase as the myocardium contracts. The partial volume effect, a useful diagnostic feature has to be accounted for during both motion estimation and temporal summation. The proposed method relies on a deformable mesh model to estimate heart motion while accounting for the partial volume effect. The estimated motion is further used to perform non-rigid summation along the time dimension. We show that the proposed method yields visual improvement on clinical data. In addition, quantitative evaluation from phantom studies proves that the proposed method achieves better noise reduction performance than available clinical techniques.