J Nucl Med. 2007; 48 (Supplement 2):446P
Technologist Abstracts
Technologist Papers I: Cardiovascular Basic, Clinical & PET |
Attenuation correction of stress PET Rb82 with ultrafast CT images
James Streeter1,
Robert Eisner1,
James Hamill2,
Michael Nelson1 and
Randolph Patterson1
1 Nuclear Cardiology, Emory Crawford Long Hospital, Atlanta, Georgia;
2 Molecular Imaging, Siemens Medical Solutions, Knoxville, Tennessee
2005
Objectives: Our study goal is to acquire aligned stress PET Rb82 and CT images as an alternative to software alignment tools, which may have large inter-operator variability and the potential to create artifactual PET results. Methods: For PET attenuation correction we acquire 3 ultrafast CT image datasets. During free breathing, each CT scan [120 kVp, 11 mAs, 0.5 sec rotation] spanned a field-of-view of 16.6 cm in 2.7 seconds with 2.0 mm thick axial slices. Dose per CT scan was 0.74 mSv. In 102 consecutive patients (pts), 1 CT (sCT1) was acquired 1-2 min after dipyridamole infusion but before acquiring stress PET, and 2 CTs (sCT2 and sCT3) were acquired post-stress PET. Sagittal, coronal and axial images were assessed for PET/CT alignment visually, and graded with the following alignment index (AI): AI=0: no apparent misalignment (ma); AI=1: small ma (1/4-1/2 myocardial wall thickness (MWTh)); AI >1 [>1/2 MWTh]. 2) On 12 randomly selected studies with AI=1, we used a 3D shift-fusion program to determine the AP, LR, and foot-to-head (z) spatial shift necessary to produce AI=0. 3) A PET normal file (nlfile) of 8 pts with low probability of CAD was established. The CT data of the 8 pts were manually shifted towards the feet (-z direction) by 1 and 2 centimeters. PET Bull’s-eyes for the 0, 1 and 2 cm shifted data were compared to nlfile and image findings were correlated with AI determination. Results: 1) 85% patients had at least 1 of 3 CTs with AI = 0, while 14% had AI=1, and 1% had AI>1. 2) Average spatial shift necessary to produce AI=0 from AI=1 was (in mm) AP: 2.8+/1.0; LR: 0.3+/-0.8; z: 4.4+/-2; total: 5.53+/-3.2. 3) Simulation study showed that 100% (9/9) studies were normal with AI=0 and AI=1 (5/5), and that 100% (10/10) studies were abnormal with AI>1. Conclusions: 1) A 3-CT protocol yields a large percentage of studies with AI= 0. 2) Average spatial shift of AI=1 studies is 5.5 +/- 3.2 mm. 3) Simulation study shows that AI=0 or AI=1 yields correct (normal) findings, while AI >1 gives incorrect (abnormal) results. 4) Using a 3-CT, ultrafast protocol, <1% of studies with AI >1 may need software alignment.
