TITLE:

Cardiac Motion Estimation in Gated Emission Computer Tomography Imagery

 

AUTHORS:

Jing Tang, W. Paul Segars, Benjamin M. W. Tsui

 

PURPOSE:

  1. Measuring the motion fields (non-rigid deformations) among gated cardiac ECT frames to characterize the cardiac function
  2. Using the motion fields among gated ECT frames as regularization constraint to improve 4D image reconstruction from projection data
  3. Summing together the motion compensated frames to reduce motion blur caused by cardiac contractile motion and therefore improve the contrast to noise properties of the ECT images

 

METHODS AND MATERIALS:

The overall estimation problem is to find a motion field consistent with elastic properties of the cardiac tissue that best matches the heart voxels in one frame to those in the other frame. The motion estimation is implemented through searching for the vector field that minimizes the cost function made of an image matching error term and a strain energy term serving as the regularization constraint. Minimizing the overall cost function is achieved by using a Taylor series expansion of the motion field and the calculus of variations on the resulting functional. Finite differencing techniques and conjugate gradient algorithm is applied to solve the resulting Euler-Lagrange equations with Neumann boundary conditions. Simulations have been performed using this motion estimation algorithm on the NCAT phantom generated images.

 

RESULTS:

Two slices of frame 1 from an 8-gate cardiac phantom are shown here with the estimated and true motion fields superimposed. As seen in the figures, the motion estimation algorithm functions reasonably well in catching the radial component of the cardiac motion but has difficulties in resolving the tangential component.

CONCLUSIONS:

More work needs to be done in tracking the cardiac twisting motion. Possible directions to go would include adding prior or using initialization with the twist involved and adding features in the cardiac wall so that twist motion is easier for the algorithm detect.

 

FUNDING SOURCES:

Corrective Image Reconstruction Methods For ECT, NIBIB

High Resolution SPECT For Molecular Imaging, NIBIB