Accuracy of motion estimation and correction in cardiac SPECT with a visual tracking system (VTS)

Objectives In this study, we investigated several possible limitations on correction accuracy of our VTS under controlled conditions using the Data Spectrum anthropomorphic phantom and Iowa heart insert.

Methods SPECT and VTS datasets of the phantom were acquired on multiple occasions. Each SPECT dataset consisted of a baseline frame-mode emission, a transmission, and a list-mode emission acquisition respectively. Motion was introduced before or during the acquisition. Synchronized motion tracking was performed during all the phases. The motion scenarios from these acquisitions included: motion between emission and transmission imaging, different % of projections corrupted by translational and rotational motion, and different timing of occurrence of motion during rotation. Both 3 degree-of-freedom (DOF) and 6 DOF motion estimation and correction were investigated. Our 3D OSEM reconstruction code was modified to automatically sense within projection motion, and reconstruct identified projections as sub-projections. The shift in center of heart counts compared to baseline and polar-map quantification were employed to estimate accuracy of correction.

Results Six DOF motion estimates successfully register corrected data to the baseline (see Table). Sub-projection reconstruction with 6 DOF estimates show a significant improvement when 18% of the projections were involved. Reconstruction with 3 DOF estimates fail to recover the shape of the heart when rotational motion was present.

Conclusions Accurate motion compensation was consistently achieved with 6 DOF estimates, however, some residual blurring due to interpolation employed in motion-correction remains.

Research Support NIH R01 EB00145

• Download figure
• Open in new tab
• Download powerpoint