Abstract
1457
Objectives Combining coronary CT angiography (CTA) and myocardial perfusion SPECT (MPS) may allow for improved diagnosis of coronary artery disease. We aimed to develop a fully automated registration scheme for combining MPS and CTA acquired at separate times on standalone scanners.
Methods We preprocess MPS and CTA to minimize the variation in voxel intensity distributions. The myocardium and blood pool on MPS is segmented and a uniform anatomical mask is used to de-emphasize influence of perfusion defects. Motion-frozen technique is used to align MPS gated data with CTA cardiac phase. Gaussian smoothing of CTA reduces intensity variations due to noise. Sum of squared differences is used as the registration cost function. Registration using the proposed method and standard mutual information (MI) based method was compared to manual alignment obtained by 2 observers on 20 stress/rest MPS/CTA datasets obtained with 2 different CT scanners (10 GE and 10 Siemens). Voxel sizes for CTA and MPS were 0.49x0.49x0.63 (mm), and 6.4x6.4x6.4 (mm).
Results Our registration of MPS with CTA had a success rate of 100%. The average registration time was 1.3±0.4 sec. MI based registration resulted in large errors (>20 mm) in all cases. The translational and rotational registration were not different from inter-observer variability (p=NS), as reported in the table. There were no differences between errors obtained for GE and Siemens datasets.
Conclusions Fully automatic registration of standalone CTA and MPS can be performed accurately in less than 2 sec and with average errors comparable to MPS voxel size.
- © 2009 by Society of Nuclear Medicine