Aarhus University Seal / Aarhus Universitets segl

Lars Poulsen Tolbod

Bayesian Penalized Likelihood Image Reconstruction (Q.Clear) in 82Rb Cardiac PET: Impact of Count Statistics

Research output: Contribution to conferenceConference abstract for conferenceResearchpeer-review

Aim: Q.Clear reconstruction is expected to improve detection of perfusion defects in cardiac PET due to the high degree of image convergence and effective noise suppression. However, 82Rb (T½=76s) possess a special problem, since count statistics vary significantly not only between patients but also during the acquisition. Image smoothening imposed by the noise penalty (beta) of Q.Clear may vary significantly between patients leading to exaggeratedly smoothened images with poor contrast in some patients and very noise images in others. In this study, we illustrate the effect of varying count statistics using a cardiac PET phantom as well as a selection of clinical patients referred for 82Rb cardiac PET. Methods: The study consistent of 3 parts: 1) A thorax-cardiac phantom was scanned for 10 minutes after injection of 1110 MBq 82Rb. Frames at 3 different times after infusion were reconstructed to yield images with total number of prompts (TP) ranging from 4 to 40 MCnts. Contrast of defect-to-LV and LV-to-cavity were characterized at a range of beta factors. 2) Body Surface Area (BSA) and TP in static frames 2.5-5 min post 82Rb infusion (1110 MBq) was extracted from 75 consecutive clinical 82Rb PET scans. 3) Static and dynamic images from a set of 7 patients (BSA: 1.6-2.2 m2) referred for 82Rb cardiac PET was analyzed using a range of beta factors. Results were compared to the institution’s standard clinical practice reconstruction protocol. All scans were performed on GE DMI Digital-Ready or D690 PET/CT. Results: 1) Contrary to the standard images, contrast of defect-to-LV and LV-to-cavity was found to depend highly on the number of TP when using Q.Clear. 2) TP in patient scans (static) could be approximated as a linear function of BSA (TP=-5e7xBSA+2e8, R2=0.7, range TP: 20-100 MCnts, BSA: 1.5-2.5 m2). 3) Optimum performance was found at beta factors ranging of 25-100. Interestingly, MBF measurements were consistent with standard protocol results within 10% at beta factors ranging from 10-100, even though, the smoothing resulting from the noise penalty varied highly from first passage (high counts, low smoothing) to the late uptake phase (low counts, high smoothing). Conclusion: The Q.Clear reconstruction algorithm can be used for the 82Rb cardiac PET. For a fixed administered 82Rb dose of 1110 MBq, the optimal range of beta factors is 25-100. However, the patient specific optimal beta factor seems to depend on both administered dose and patient BSA.
Original languageEnglish
Publication year2017
Publication statusPublished - 2017
EventEuropean Association of Nuclear Medicine Annual Congress - Vienna, Austria
Duration: 21 Oct 201725 Oct 2017


ConferenceEuropean Association of Nuclear Medicine Annual Congress
Internet address

See relations at Aarhus University Citationformats

ID: 120303026