Search

Background Transmission (TX) scan time by use of radionuclide sources for cardiac positron emission tomography prolong imaging and increase the likelihood of patient motion artifacts. A reconstruction algorithm combining ordered-subsets expectation maximization with a Bayesian prior was developed and applied to rapid Germanium-68 (Ge-68) TX scans. Methods and Results A cardiac phantom with Fluorine-18 (Fl-18) was used to determine a minimal count threshold for Ge-68 TX scanning. Images were acquired over a count range from 2.5 × 106 to 8 × 107 and for a high-count scan of 1.6 × 109 counts to study reconstruction parameters and to determine the minimum TX count threshold. The method was compared against clinical 4-minute TX scans in ten Rubidium-82 (Rb-82) rest/stress myocardial perfusion studies (body mass index, 30 ± 4 kg/m2). The minimal count threshold was 20 × 106, and the mean scan time for the Rb-82 studies was 70.5 ± 3.4 seconds. More than 90% of the segmental scores computed from images acquired via rapid TX scans differed by less than 5% from those obtained with 4-minute TX scans. The mean differences in perfusion scores between the rapid and 4-minute TX scans were 0.46% (95% confidence interval, −1.84% to 0.93%) at rest and 0.39% (95% confidence interval, −1.84% to 1.07%) at stress, demonstrating equivalency of the rapid and 4-minute scans. Conclusions Ordered-subsets expectation maximization with a Bayesian prior accurately and efficiently reconstructs rapidly acquired Ge-68 TX scans for Rb-82 myocardial perfusion positron emission tomography studies.

High-quality attenuation maps are critical for attenuation correction of myocardial perfusion single photon emission computed tomography studies. The filtered backprojection (FBP) approach can introduce errors, especially with low-count transmission data. We present a new method for attenuation map reconstruction and examine its performance in phantom and patient data.The Bayesian iterative transmission gradient algorithm incorporates a spatially varying gamma prior function that preferentially weights estimated attenuation coefficients toward the soft-tissue value while allowing data-driven solutions for lung and bone regions. The performance with attenuation-corrected technetium 99m sestamibi clinical images was evaluated in phantom studies and in 50 low-likelihood patients grouped by body mass index (BMI). The algorithm converged in 15 iterations in the phantom studies. For the clinical studies, soft-tissue estimates had significantly greater uniformity of mediastinal coefficients (mean SD, 0.005 cm(-1) vs 0.011 cm(-1); P.0001). The accuracy and uniformity of the Bayesian iterative transmission gradient algorithm were independent of BMI, whereas both declined at higher BMI values with FBP. Attenuation-corrected perfusion images showed improvement in myocardial wall variability (4.8% to 4.1%, P = .02) for all BMI groups with the new method compared with FBP.This new method for attenuation map reconstruction provides rapidly converging and accurate attenuation maps over a wide spectrum of patient BMI values and significantly improves attenuation-corrected perfusion images.

Background: Ge-68 rod sources are utilized in PET scanners for measuring non-uniform patient attenuation. To compensate for insufficient signal in larger patients or shortened acquisition time, attenuation maps (AM) are segmented prior to attenuation correction. Although segmented FBP (SFBP) is widely utilized, it can be of limited value when the projection data has significant noise. We investigate a previously validated Bayesian reconstruction method (BITGA) from cardiac SPECT to improve the accuracy of AM estimates in low count PET studies acquired with significantly reduced scan time. Methods: BITGA uses a prior function that preferentially weights the current attenuation coefficient estimate at each pixel toward the value for soft-tissue region. Lung, air and bone regions are minimally weighted allowing a data-driven solution. The prior function was optimized for the soft-tissue (water) narrow-beam linear attenuation coefficient at 511 keV (0.097 cm-1). The data was created from smoothing and normalizing projection data to a high-count reference scan. Ge-68 transmission scans of 1 and 4-minute of a female anthropomorphic phantom were acquired on a CTI ACCEL™ LSO PET scanner and reconstructed by FBP, SFBP and BITGA. We examined profiles across the phantom chest and calculated the mean difference of attenuation coefficients (MDAC) between both scans. A 2-minute scan of patient data was studied to evaluate the noise-reduction capability of BITGA in a large patient (BMI=49). Results: In the phantom study, the 1-minute scan increased approximately by 2 times the projection noise components of the 4-minute scan. The MDACs of 1-minute scan were 0.020/cm (FBP), 0.0038/cm (SFBP), 0.0018/cm (BITGA) (p Conclusions: BITGA is a promising new method for accurately reconstructing the Ge-68 AM acquired with significantly reduced time for cardiac PET compared with the current methods of FBP and SFBP. Download : Download full-size image

Background: Tc99m-Teboroxime (TEBO) has been pursued as a myocardial perfusion agent because of its high extraction fraction. However, to image TEBO accurately on traditional SPECT is difficult due to a fast washout rate and the high hepatic uptake. In this work, we develop and demonstrate a new attenuation correction (AC) technique for dynamic SPECT to improve TEBO's perfusion distribution using a novel energy-based listmode scatter correction (SC) combined with AC. Methods: We recruited five subjects having previous clinically indicated rest-stress AC Tc-99m-MIBI SPECT to receive an adenosine stress TEBO. The Gd-153 transmission study was acquired sequentially prior to the TEBO study on Siemens ECAMTM system modified to acquire listmode data in a rapid phased fanning mode. 10s per phase with 32 phases and 32 projection angles in each phase were acquired in 180o RAO-LPO arc during peak adenosine stress. Listmode data were rebinned into energy spectra of projections. The spectrum at each pixel was corrected using a previously reported energy-based SC technique (IESD). Attenuation maps were reconstructed with a Bayesian iterative approach (BITGA) and AC was performed with MLEM algorithm using the IESD corrected projections. Segmental difference from the polar map samplings (total 90 segments) of non-corrected TEBO (NC) TEBO to AC MIBI were compared with AC TEBO to AC MIBI for assessing the improvements of patients' perfusion distribution after AC. Results: The mean segmental difference between NC TEBO and AC MIBI was 7.5 and reduced to 6.0 after AC (p Conclusion: AC combined with IESD correction demonstrates improved perfusion distribution of TEBO by reducing 20% (p Download : Download full-size image

Objectives: LSO-based PET detectors have high-count rate capability with the potential to greatly improve the quality and efficiency of Rb-82 cardiac PET. Although 2D myocardial perfusion PET has been well-established, many new PET and PET-CT scanners only have 3D imaging capability. The purpose of this study was to quantitatively compare 2D and 3D Rb-82 myocardial perfusion images in normal and fixed defect populations. Methods: A retrospective analysis of the resting component from rest/stress Rb-82 myocardial perfusion PET studies was performed in 40 patients (20 normal, 20 fixed defect). The dynamic rest 2D and ECG-gated rest 3D images were acquired sequentially on a Siemens ECAT Accel™ LSO scanner. All studies were reconstructed using OSEM (6 it., 8 subsets, Butterworth filter 5th order, 12 mm kernel). Segmental raw counts (normalized to the max pixel) were quantified from the short axis images after porting to the QPS™ processing environment. The 2 apical segments were excluded. For the fixed defect group, an independent criteria for segmental abnormality was derived from the 20 normal studies as > 2 SD below the mean segmental value. Paired t-tests were used to measure significance of differences between 2D and 3D values. Results: Mean segmental scores in the normal group were highly equivalent, 85.7±5.4 and 85.6±5.9 for 2D and 3D acquisitions respectively (p=0.65). 336/360 segments (93%) had Conclusion: 2D and 3D Rb-82 PET images are statistically identical for normal and abnormal segments. Defect location was also identical for the 2D and 3D methods. This suggests that 3D Rb-82 perfusion imaging is identical to validated 2D methods with the added value of ECG-gating.

Introduction: Filtered backprojection (FBP) reconstructions for transmission (Tx) data can be used for attenuation correction (AC) in phantom data, however impact on body mass index (BMI) has not been adequately demonstrated. We performed a paired comparison of Tx maps and AC myocardial perfusion images (Em) in patients with clinically relevant range of BMI (20–45 kg/m2) processed with the Bayesian Iterative Transmission Gradient Algorithm (BITGA) and FBP. Methods: We examined 50 rest/stress Tc-99m Sestamibi patient studies (5 male and 5 female in five BMI groups: 40 kg/m2). with a low-likelihood ( Data Acquisitions: Data was acquired on a Philips Cardio system using VantagePro™ and ExSPECT II™ software using LEHR collimation, 30s/25s (rest/stress) per stop, 64 angles. Results: Tx maps appeared less accurate and less uniform when reconstructed with FBP (see figures). For Em data, BITGA reduced the wall to wall variability from 4.8% (BITGA) to 4.1% (FBP) (p=0.02). Furthermore, the number of patients whose septal-lateral wall ratios (SL) differed from unity by more than ten percent was reduced from 12% to 4% (p 30 were more likely to have a divergent SL (83.3% vs 16.7%). Visual inspection of FBP based images indicate this may be due to poor definition of the lung/heart boundary. Download : Download full-size image Conclusion: Use of BITGA produced more accurate and uniform TX results for all BMI groups. Also, patients with BMI>30, use of FBP Tx maps introduced artifacts that can compromise Em quality.