Comparison of 2D, 3D high dose and 3D low dose gated myocardial 82Rb PET imaging
© Knešaurek et al.; licensee BioMed Central Ltd. 2007
Received: 28 March 2007
Accepted: 22 October 2007
Published: 22 October 2007
We compared 2D, 3D high dose (HD) and 3D low dose (LD) gated myocardial Rb-82 PET imaging in 16 normal human studies. The main goal in the paper is to evaluate whether the images obtained by a 3D LD studies are still of comparable clinical quality to the images obtained with the 2D HD or 3D HD studies.
All 2D and 3D HD studies were performed with 2220 MBq of Rb-82. The 3D LD were performed with 740 MBq of Rb-82. A GE Advance PET system was used for acquisition. Polar maps were created and used to calculate noise among (NAS) and within (NWS) the segments in the noise analysis. In addition, the contrast between left ventricular (LV) wall and LV cavity was also analysed. For 13 subjects, ejection fraction (EF) on 2D and 3D studies was calculated using QGS program.
For the H20 reconstruction filter, the mean contrast in mid-ventricular short-axis slice was 0.33 ± 0.06 for 2D studies. The same contrast for the 3D HD studies was 0.38 ± 0.07 and for 3D LD, it was 0.34 ± 0.08. For the 6 volunteers where 3D HD was used, NAS was 3.64*10-4 and NWS was 1.79*10-2 for 2D studies, and NAS was 3.70*10-4 and NWS was 1.85*10-2 for 3D HD studies, respectively. For the other 10 volunteers where 3D LD was used, NAS was 3.85*10-4 and NWS was 1.82*10-2 for the 2D studies, and NAS was 5.58*10-4 and NWS was 1.91*10-2 for the 3D LD studies, respectively. For the sharper H13 filter, the data followed the same pattern, with slightly higher values of contrast and noise. EF values in 2D and 3D were close. The Pearson's correlation coefficient was 0.90. The average difference from 13 subjects was 8.3%.
2D and 3D HD gating Rb-82 PET cardiac studies have similar contrast, ejection fractions and noise levels. 3D LD gating imaging, gave comparable results in terms of contrast, EF and noise to either 2D or 3D HD gating PET imaging. 3D LD PET gated imaging can make Rb-82 PET cardiac imaging more affordable with significantly less radiation exposure to the patients.
Not long ago , we compared 2D with 3D modes in myocardial 82Rb PET imaging at rest. Here, we would like to extend the same comparison to gating myocardial 82Rb PET imaging at rest. The gating imaging provides additional useful information like ejection fraction (EF) and wall thickening. However, it is more demanding due to loss of counts (e.g. bad beats rejection) and dynamic memory limitations. Due to the short half-life of 82Rb (75 s), 82Rb PET cardiac images tend to be count-poor. Additional shifting of counts in different hearth cycle phases makes gating 82Rb PET myocardial imaging even more challenging than non-gating imaging. Also, 82Rb biokinetics, i.e., high blood pool activity approximately 2 min after I.V. injection, combined with 82Rb short half-life, requires a careful acquisition protocol in order to obtain images of adequate quality. While 2D and 3D 18F – Fluorodeoxyglucose (FDG) imaging can be optimized based on PET system performance characteristics, mostly described by a noise equivalent count (NEC) rate , the dynamic 82Rb PET cardiac imaging is more complicated. Optimization of 82Rb PET imaging requires taking into account 82Rb biokinetics and 82Rb short half-life, in addition to the PET system performance characteristics.
The main goal in the paper is to evaluate whether the images obtained by a low dose (LD) of 740 MBq (20 mCi) in the 3D myocardial 82Rb perfusion gated PET studies are still of comparable clinical quality to the images obtained with the high dose (HD) of 2220 MBq (60 mCi) in 2D and 3D PET 82Rb perfusion PET gated studies. The reduction in dose by a factor of three has significance in reducing costs associated with 82Rb and the consequent potential of making 82Rb perfusion PET myocardial imaging more affordable. Reducing the patient dose by a factor of three also significantly reduces exposure to the patients.
All 2D volunteer studies were performed by injecting I.V. 2220 MBq (60mCi) of 82Rb. For six volunteers, 3D studies were performed with a high dose (HD) of 2220 MBq of 82Rb and for 10 volunteers in the 3D studies, a low dose (LD) of 740 MBq (20 mCi) of 82Rb was used. One volunteer participated twice, i.e., was imaged at two separate occasions, in 2D and 3D HD, and 2D and 3D LD studies, respectively. In all studies, i.e., 2D and 3D LD and 3D HD studies, time per frame was 1 minute and total acquisition time after appropriate delay, was 6 minutes. In the 2D studies there was a delay of 2 minutes, in the 3D LD studies there was a delay of 3 minutes, and in the 3D HD studies there was a delay of 5 minutes. The heart cycle was divided in 8 phases in the 2D gated PET studies. In the 3D gated PET studies, 5 phases were used, due to dynamic memory limitations. These human protocols were approved by the Institutional Review Board of our Medical Center.
The GE ADVANCE (General Electric Medical Systems, Milwaukee, WI) system was used for all acquisitions in both the 2D and 3D modes. The 2D images were reconstructed using a filtered backprojection reconstruction method and Hanning filters with a 0.33 cycles/pixel (H13) and 0.21 cycles/pixel (H20), cutoff frequency, respectively. The 3D studies were reconstructed using a Kinahan-Rogers  algorithm and also using H13 and H20 filters. The matrix size was 128 × 128 and the pixel size was 4.29 mm. Attenuation correction using an 8-min transmission scan was applied in all studies. In the 2D studies, Bergstrom  scatter correction was applied. For the 3D data, scatter correction was performed by fitting the tails of the sinogram to a 2D Gaussian . Transaxial gated slices were transferred to GE Xeleris system for further gated analysis.
Intrasegmental variance was investigated by calculating variance for each segment, as well as all segmental variances and the average coefficient of variance. F statistics were used to compare 2D and 3D HD studies, between 2D and 3D low dose studies and 3D HD and 3D low dose studies.
Intersegmental variance was investigated by applying a logarithmic transform on each mean segmental value and performing two-way ANOVA without replication. The effects of different patients, different segmental positions and interaction of different patients and segments (assumed none) on noise were tested. Again, F statistics were used to compare 2D and 3D HD studies, between 2D and 3D low dose studies and 3D HD and 3D low dose studies.
EFs were calculated using QGS program. Initially we did not plan to compare EF in 2D and 3D studies due to difference in the number of phases used to cover the heart cycle. As mentioned before, because of the dynamic memory limitations, only 5 phases were used in 3D gated studies. In retrospective comparison between 2D and 3D EFs, 3 sets of data were found to be corrupted and only 13 subject were used in the comparison. Also, our noise and contrast analysis showed that the 3D LD and 3D HD images were very comparable. Therefore, we did not distinguish between HD and LD 3D studies in the comparison between 2D and 3D EFs. A second reason for not splitting between HD and LD 3D studies was the limited number of subjects and we want to keep the same number of studies in 2D and 3D.
Comparison of contrast values for 2D and 3D high-dose (HD) and 3D low-dose (LD) in 82Rb PET myocardial imaging -smoother filter H20
NAS × 10-4
NWS × 10-2
Comparison of contrast values for 2D and 3D high-dose (HD) and 3D low-dose (LD) in 82Rb PET myocardial imaging – sharper filter H13
NAS × 10-4
NWS × 10-2
Table 1 and table 2 gives the mean contrast values in mid-ventricular short-axis slice for 2D and 3D low and high dose studies for H20 and H13 reconstruction filters, respectively. The contrast values in 2D and 3D LD studies are very close. The contrast values in 2D and corresponding 3D HD studies are also close, with 3D HD studies having slightly higher values. The p value of 0.04 shows that there was no statistically significant correlation between noise in 3D low and high dose studies. The contrast values are higher for the H13 reconstruction filter than for the smoother H20 (table 1 and table 2). Noise among segments (NAS) and noise within segments (NWS) are also given for 2D and 3D low and high dose studies, for both filters. From table 1 and table 2 one can see that noise among and within the segments is higher for H13, i.e., for sharper filter than for H20 in 2D and 3D low and high dose studies. Noise among and within the segments were very similar for the 2D and 3D high dose (HD) studies, although there were no statistically significant correlations between noise within segments in 2D and 3D HD studies (p = 0.02). In comparison between 2D and 3D low dose (LD) studies, noise among and within the segments was moderately higher for the 3D LD studies. However, again there were no statistically significant correlations between noise among and within segments in 3D LD studies and 2D and 3D HD studies (p = 0.025 for H20 filter and p = 0.001 for H13 filter). For the sharper filter H13, noise within segments was higher than for the smoother filter H20 and was practically the same in 2D and 3D low and high dose studies. The H13 filter provided relatively high noise images and is not used routinely in clinical practice.
Gated myocardial images are usually count poor. Gated 82Rb PET myocardial imaging is an even greater challenge, because 82Rb has quite a short half-life, making 82Rb images count poor even without gating. Therefore, 3D gated 82Rb PET myocardial imaging has great appeal due to significantly higher sensitivity than 2D imaging. The first goal of our study was to investigate whether gated 3D high dose (2220 MBq) imaging, i.e., the same dose as in gated 2D images, would achieve better performance due to higher sensitivity. The results of our study indicate that gated 3D high dose images did not provide better images, because of the longer delay in acquisition from the time of injection in 3D HD imaging, as discussed below. The longer delay in 3D HD imaging diminishes the advantages of the higher sensitivity in 3D imaging in comparison with 2D imaging. The second question was to investigate the possibility of replacing high dose gated 2D and gated 3D imaging with gated 3D low dose (740 MBq) imaging. The advantage of 3D LD imaging, due to 3D mode higher sensitivity, is the same count rate as in 2D HD mode, but with a lower injected dose. This could lead to significant cost savings in the purchase of an 82Rb generator and thus could make myocardial 82Rb PET imaging more affordable .
Our results show that the contrast values between LV and LV cavity were practically the same in 2D and 3D low and high dose studies. However, the noise in 3D low dose studies has been slightly higher when compared with the 2D and 3D high dose studies. Nevertheless, in spite of the slight increase in noise in the 3D LD studies, the images are very comparable with high dose 2D and 3D images. Due to the fact that we used 8 phases in the 2D studies and only 5 phases in the 3D studies, the ejection fraction values in 3D studies were slightly underestimated, in average by 8.3%. The same effect was observed in the comparison between 16 and 8 phase gated SPECT studies, where the 8 phase studies show 3.71% lower ejection fractions .
In terms of dosimetry, the effective dose equivalent for 2220 MBq (60mCi) of 82Rb is 2.66 mSv. The kidneys, as the critical organ, receives a dose of 19.98 mGy . For 740 MBq (20 mCi) of 82Rb, the effective dose equivalent is 0.89 mSv with the kidneys receiving only 6.66 mGy, one-third of the dose for 2220 MBq (60mCi).
On our dedicated high counting-rate performance PET system, 3D high dose (2220 MBq) gated PET imaging gives similar contrast and noise level as high dose 2D imaging. However, high dose 3D gated imaging did not achieve a better performance due to a necessary delay in acquisition from the time of injection, and slightly higher randoms and scatter fraction. Low dose (720 mBq) 3D gated imaging, while achieving similar contrast and ejection fractions, resulted in slightly higher noise, compared to either 2D or high dose 3D imaging. In view of these findings, we conclude that 3D low dose acquisition images with optimized filtering can probably give acceptable results with significant cost savings, related to purchasing an 82Rb generator, and considerable decrease in patient exposure.
- Knešaurek K, Machac J, Krynyckyi BR, Almeida OD: Comparison of 2D and 3D82Rb myocardial perfusion PET imaging. J Nucl Med. 2003, 44: 1350-1356.PubMedGoogle Scholar
- Lartizien C, Comtat C, Kinahan PE, Ferreira N, Bendriem B, Trébossen R: Optimization of the injected dose based on noise equivalent count rates for 2- and 3-dimensional whole-body PET. J Nucl Med. 2002, 43: 1268-1278.PubMedGoogle Scholar
- Kinahan PE, Rogers JG: Analytic three dimensional image reconstruction using all detected events. IEEE Trans Nucl Sci. 1989, 36: 964-968. 10.1109/23.34585.View ArticleGoogle Scholar
- Bergstrom M, Eriksson L, Bohm C, Blomqvist G, Litton J: Correction for scattered radiation in a ring detector positron camera by integral transformation of the projections. J Comput Assist Tomogr. 1983, 7: 42-50. 10.1097/00004728-198302000-00166.View ArticlePubMedGoogle Scholar
- Stearns CW: Scatter correction method for 3-D PET using 2-D fitted Gaussian functions [abstract]. J Nucl Med. 1995, 36: 105-Google Scholar
- Germano G, Kiat h, Kavanagh PB, Moriel M, Mazzanti M, Su HT, surVan Train KF, Berman DS: Automatic quantification of ejection fraction from gated myocardial perfusion SPECT. J Nucl Med. 1995, 36: 2138-2147.PubMedGoogle Scholar
- Knešaurek K: New developments in PET instrumentation. Quo Vadis PET ?. J Nucl Med. 2001, 42: 1831-1832.PubMedGoogle Scholar
- Ollinger JM: Model-based scatter correction for fully 3D PET. Phys Med Biol. 1996, 41: 153-176. 10.1088/0031-9155/41/1/012.View ArticlePubMedGoogle Scholar
- Stabin MG, Stubbs JB, Toohey RE: Radiation dose estimates for radiofarmaceuticals. Radiation Internal Dose Information Center, Oak Ridge Institute for Science and Education, Oak Ridge, TN. 1996Google Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2385/7/4/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.