Kirsten Bouchelouche

Static Myocardial Perfusion Imaging using denoised dynamic Rb-82 PET/CT scans

Publikation: KonferencebidragPosterForskningpeer review

Introduction: Relative and absolute measures of myocardial perfusion are derived from a single 82Rb PET/CT scan. However, images are inherently noising due to the short half-life of 82Rb. We have previously shown that denoising techniques can be applied to dynamic 82Rb series with excellent quantitative accuracy. In this study, we examine static images created by summing late frames of denoised dynamic series. Method: 47 random clinical 82Rb stress and rest scans (27 male, age 68+/- 12 y., BMI 27.9 +/- 5.5 kg/m2) performed on a GE Discovery 690 PET/CT scanner were included in the study. Administered 82Rb dose was 1110 MBq. Denoising using HYPR-LR or Hotelling 3D algorithms was performed as post-processing on the dynamic images series. Static series were created by summing frames from 2.5-5 min. The image data was analysed in QPET (Cedars-Sinai). Relative segmental perfusion (normalized to the hottest pixel) was compared in order to assess whether a correlation between original image data and denoised image data could be found. In addition to this, correlations for TPD, Extent of defect and summed defect scores (SSS, SRS and SDS) were investigated. The data was analysed using linear regression and Bland-Altman analysis. Results: For HYPR-LR, a good correlation was found for relative segmental perfusion for both stress (y=1.007x+0.313, R2=0.98) and rest (y=1.007x+ 0.421, R2=0.96) scans with negative bias of -0.79±1.44 and -0.90±1.63, respectively. Correlations for SSS (R2=0.94), SRS (R2=0.92), SDS (R2=0.89), TPD (stress, R2=0.99; rest, R2=0.97) and Extent (stress, R2=0.99; rest, R2=0.96) were likewise excellent. For Hotelling 3D, correlations were slightly lower than for HYPR-LR: relative perfusion (rest, y=1.016x-0.131, R2=0.95; stress, y=0.982x+3.165, R2=0.87), SSS (R2=0.90), SRS (R2=0.88), SDS (R2=0.75), TPD (stress, R2=0.98; rest, R2=0.93) and Extent (stress, R2=0.97; rest, R2=0.90).Conclusion: Static 82Rb series for quantifying relative myocardial perfusion can be created by summing denoised late frames of dynamic 82Rb series. Excellent quantitative accuracy was found for both HYPR-LR and Hotelling 3D algorithms. This enables either an improvement in image quality or a reduction in administered dose. Further studies are required to define the optimal tradeoff between administered dose and image quality.
Udgivelsesår10 okt. 2015
StatusUdgivet - 10 okt. 2015
BegivenhedEANM'15 - Hamborg, Hamborg, Tyskland
Varighed: 11 okt. 201514 okt. 2015



Se relationer på Aarhus Universitet Citationsformater

ID: 93625672