Automatic registration of misaligned CT attenuation correction maps in Rb-82 PET/CT improves detection of angiographically significant coronary artery disease

Background. We aimed to evaluate the utility of fully automated software registration intended to improve CT attenuation correction (CTAC) map misalignments during cardiac Rb-82 PET/CT myocardial perfusion imaging (MPI). Methods. 171 consecutive patients (108 males, mean age 69 years), undergoing both rest-stress Rb-82 PET/CT MPI and invasive coronary angiography within 6 months (mean 14 days, range 0-170), were studied. List mode data were automatically processed in batch mode to generate transaxial attenuation corrected slices with four different CTAC alignment correction strategies: (i) no alignment correction (NONE); (ii) manual correction (MANUAL); (iii) automated 6-parameter rigid correction (AUTO); and (iv) targeted use of automated correction only where PET-CTAC alignment was initially judged as incorrect on either stress or rest scan (AUTO for misalignment only). Initial and final registration quality was graded (1-3) by an experienced radiologist (1: satisfactory alignment (< 2 mm misalignment), 2: slight misalignment (2-5 mm in any direction), or 3: poor (> 5 mm misalignment in any direction). Total perfusion deficit (TPD) and ischemic TPD (ITPD) were computed automatically, and their diagnostic accuracy to detect significant coronary artery disease with each realignment technique was assessed using receiver operating characteristic analysis. Results. The diagnostic accuracy of ITPD, expressed as area under curve, was .81 +/- A .03 with no alignment correction (NONE), .83 +/- A .03 with MANUAL correction, .85 +/- A .03 with AUTO correction (P < .05 vs. NONE and MANUAL), and .87 +/- A .03 with the targeted use of AUTO correction (P < .05 vs. NONE, MANUAL and AUTO). Both manual and software corrections increased the percentage of cases with satisfactory PET-CTAC map alignment (P < .05 for all) at rest (from 55% for NONE to 80% for MANUAL and 92% for AUTO) and at stress (from 51% for NONE to 78% for MANUAL and 84% for AUTO). Conclusion. The diagnostic accuracy of Rb-82 PET/CT MPI with automated rigid alignment is improved compared to data with no CTAC scan alignment or with manual alignment. The optimal strategy for diagnostic performance is to apply automatic alignment only in cases which are visually identified as misaligned.