New methods for registering long-time I-123 ADAM SPECT image sequences to magnetic resonance images.

摘要

BACKGROUND: I-123 ADAM single-photon emission tomography (SPECT) analysis is usually performed by drawing regions of interest or by directly registering the SPECT image to a magnetic resonance image (MRI). Each method is limited by an observer or registration bias. METHODS: Eleven healthy volunteers were recruited for I-123 ADAM SPECT using a triple-headed gamma camera with fan-beam collimators after an intravenous injection of 185 MBq of I-123 ADAM. We propose two new registration methods: sequential registration and a two-step registration. Reconstructed images were registered to MRIs using mutual information with sequential and two-step methods. RESULTS: The sequential and two-step method mean counts of midbrain activity were not significantly different (166.1+/-4.3 vs. 164.2+/-4.1, P=0.273), but they were significantly higher than the direct method mean count (132.8+/-3.7, P<0.001). Sequential and two-step method midbrain/cerebellum ratios were not significantly different (4.40+/-0.37 vs. 4.28+/-0.34, P=0.277), but they were significantly higher than the direct method midbrain/cerebellum ratio (1.68+/-0.16, P<0.001). CONCLUSION: The accuracy of the sequential method of registration was significantly better than that of drawing regions of interest or directly coregistering the SPECT image to an MRI. It is, however, tedious and time-consuming. The simplified two-step method yields similar results and is more practical for clinical I-123 ADAM SPECT studies.