Optimisation of reconstruction--reprojection-based motion correction for cardiac SPECT

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Cardiology

Background:

• Cardiac motion significantly degrades myocardial perfusion SPECT images, causing artifacts.
• Reconstruction-reprojection algorithms are effective for automatic motion correction.
• Previous methods lacked full optimization of parameters and implementation details.

Purpose of the Study:

• To optimize and compare two reconstruction-reprojection based motion correction techniques for myocardial perfusion SPECT.
• To evaluate the impact of different cost functions and iteration numbers on motion correction accuracy and efficiency.

Main Methods:

• Two methods were compared: Method 1 (projection space correction) and Method 2 (reconstruction space correction).
• Optimized parameters included cost functions (squared difference, normalized cross-correlation, mutual information) and iteration count.
• Methods were tested on simulated cardiac perfusion SPECT data with added lateral shift, vertical shift, and vertical creep.

Main Results:

• Method 2 showed slightly better overall performance but had significantly longer execution times than Method 1.
• Mutual information consistently yielded the best results across all motion types and methods.
• Three iterations were sufficient for Method 1 to achieve good quality correction.

Conclusions:

• The traditional reconstruction-reprojection method (Method 1) with mutual information and three iterations is a clinically viable option for myocardial perfusion SPECT motion correction.
• While reconstruction-space correction (Method 2) offers marginal benefits, its computational cost limits its clinical applicability.