Jørgen Frøkiær

Objectives: Absolute quantification of myocardial blood flow (MBF) using 82Rb has shown promising results for diagnosing coronary artery disease. The short half-life of 82Rb, however, requires a high administered dose in order to get sufficient counting statistics at later timepoints. This increases radiation dose to the patient, reduces Sr/Rb generator lifetimes and increases risk of scanner saturation. The aim of this study was to evaluate quantitative performance of several image denoising techniques which could ultimately be used to reduce administered activity.

Methods: Fifty patients with suspected ischemic heart disease underwent a dynamic 7 minute 82Rb scan under resting and adenosine induced hyperaemic conditions after injection of 1100 MBq of 82Rb on a GE Discovery 690 PET/CT. Dynamic images were filtered using HighlY constrained backPRojection (HYPR) and a Hotelling filter of which the latter was evaluated using a range of 4 to 7 included factors and for both 2D and 3D filtering. Data were analyzed using Cardiac VUer and obtained MBF values were compared with those obtained when no denoising of the dynamic data was performed.

Results: Both HYPR and Hotelling denoising could be performed successfully for all scans. Quantitative performance was suboptimal when 5 or less factors were used for the Hotelling method, both in 2D and 3D. For 6 or more factors and for HYPR-LR, excellent quantitative accuracy was obtained when compared to non-denoised data (r2 = 0.996, slope = 1.013 for 2D Hotelling with 6 factors; r2 = 0.994, slope = 0.998 for 3D Hotelling with 6 factors; r2 = 0.995, slope = 0.996 for HYPR-LR).

Conclusions: Both HYPR-LR and Hotelling denoising methods can be applied to 82Rb data with excellent quantitative accuracy. This enables an improvement in image quality or potentially a significant reduction in administered dose.