Attenuation Correction for Stress and Rest PET ⁸²Rb Myocardial Perfusion Images

TO THE EDITOR: The July 2007 article by Gould et al. (1) reported a 40% false-positive rate for cardiac PET ⁸²Rb myocardial perfusion imaging with CT attenuation correction, using helical slow imaging (29 s) during free breathing and helical fast imaging (4 s) during a breath-hold at end expiration. Further, the authors suggested that correction with nonhelical, time-averaged cine CT images eliminates artifactual defects in PET ⁸²Rb images. Our concern with the study is that it contrasts false-positive findings from cine CT with software alignment and false-positive findings from helical CT without software alignment. Because most manufacturers of PET/CT scanners have a software alignment tool to be used in conjunction with helical CT, we suggest that it is appropriate and important for Gould et al. to compare false-positive findings from software-aligned cine CT and false-positive findings from software-aligned helical CT (e.g., their slow helical CT scan). Table 5 of Gould et al. lists an artifact frequency for unshifted slow helical CT studies (27%, or 39/145) similar to that found for “conventional” PET ⁸²Rb studies (21%, or 252/1,177) in an earlier publication by Dr. Gould's group (2), in which a ⁶⁸Ge rod source was used for the attenuation correction. After visual checking of the PET ⁸²Rb and attenuation images for misregistration, the misaligned conventional rod source studies were manually shifted using computer software (2). Moreover, even with cine CT, Gould et al. reported that 19% (22/114) of patient datasets were misaligned with PET and required software alignment (1). That is, the new paper (1) combined with the earlier publication (2) supports the conclusion that free-breathing helical CT and cine CT have nearly the same frequency of artifacts as does conventional PET. Significantly, all techniques required a software alignment solution.

In our institution, we have used a somewhat different approach on our 64-slice Biograph PET/CT scanner (Siemens). We acquire 3 very fast (2.7-s) helical CT scans during free breathing, 1 immediately before and 2 immediately after acquiring the stress PET ⁸²Rb images (3). The exposure, CTDI_vol, is 0.7 mGy in each scan. We have found that this protocol increases the probability of alignment between a PET ⁸²Rb image and an acquired CT image. Our protocol also includes 3 CT scans at rest for correction of the rest PET ⁸²Rb scan. Like Gould et al. (1), we have estimated visually the degree of PET/CT misalignment with this procedure using the PET/CT 3-dimensional fusion software of the manufacturer (3). We found no apparent misalignment between PET and at least 1 CT scan in 85% of studies at stress and 89% of studies at rest. The best-case misalignment was small, and appropriate for PET attenuation correction, in an additional 14% of the studies at stress and 11% of the studies at rest (3). In only a few cases (<1%) did we observe a large or severe PET misalignment with all 3 of the CT scans that then required computer software alignment. We have acquired 1,400 rest/stress PET/CT ⁸²Rb clinical studies with this protocol. The total CTDI_vol is 4.2 mGy with our 6–CT scan protocol. Gould et al. quoted a radiation exposure of 5.7 mGy for their helical CT scan and a radiation dose of 10 mGy for cine CT. We are studying techniques to reduce the dose even further. These steps include reducing the x-ray voltage from 120 to 100 kVp and even to 80 kVp in very thin patients and reducing the number of CT scans, thus requiring a greater reliance on software alignment.

In summary, the slow helical non–breath-hold CT approach originally proposed by Brunken et al. (4) produces a frequency of misalignment-related artifacts that is similar to the frequency reported for cine CT (1) and conventional PET (2). Gould et al. (1) did not provide the false-positive rate for software-shifted non–breath-hold helical CT, and this omission represents a major limitation of the paper. PET and CT alignment can be achieved with a fast helical multi–CT scan protocol that limits the need for software alignment tools to a small percentage of studies, while using an even lower radiation dose (3).