Optimization of Q.Clear reconstruction for dynamic 18F PET imaging

EJNMMI Physics

Original Research

Kyrre Eeg Emblem¹  Lars Tore Gyland Mikalsen²  Elisabeth Kirkeby Lysvik³  Trine Hjørnevik³  Mona-Elisabeth Rootwelt-Revheim⁴ 

[1] Department of Physics and Computational Radiology, Division of Radiology and Nuclear Medicine, Oslo University Hospital, Building 20, Gaustad Sykehus, Sognsvannveien 21, 0372, Oslo, Norway;Department of Physics and Computational Radiology, Division of Radiology and Nuclear Medicine, Oslo University Hospital, Building 20, Gaustad Sykehus, Sognsvannveien 21, 0372, Oslo, Norway;Department of Life Sciences and Health, Oslo Metropolitan University, Oslo, Norway;Department of Physics and Computational Radiology, Division of Radiology and Nuclear Medicine, Oslo University Hospital, Building 20, Gaustad Sykehus, Sognsvannveien 21, 0372, Oslo, Norway;Institute of Clinical Medicine, University of Oslo, Oslo, Norway;Institute of Clinical Medicine, University of Oslo, Oslo, Norway;The Intervention Centre, Oslo University Hospital, Oslo, Norway;Department of Nuclear Medicine, Division of Radiology and Nuclear Medicine, Oslo University Hospital, Oslo, Norway;

关键词: Dynamic PET;    Quantitation;    Recovery coefficient;    β-factor;    Q.Clear;   

DOI  :  10.1186/s40658-023-00584-1

 received in 2023-06-19, accepted in 2023-10-12,  发布年份 2023

来源: Springer

【 摘 要 】

BackgroundQ.Clear, a Bayesian penalized likelihood reconstruction algorithm, has shown high potential in improving quantitation accuracy in PET systems. The Q.Clear algorithm controls noise during the iterative reconstruction through a β penalization factor. This study aimed to determine the optimal β-factor for accurate quantitation of dynamic PET scans.MethodsA Flangeless Esser PET Phantom with eight hollow spheres (4–25 mm) was scanned on a GE Discovery MI PET/CT system. Data were reconstructed into five sets of variable acquisition times using Q.Clear with 18 different β-factors ranging from 100 to 3500. The recovery coefficient (RC), coefficient of variation (CVRC) and root-mean-square error (RMSERC) were evaluated for the phantom data. Two male patients with recurrent glioblastoma were scanned on the same scanner using 18F-PSMA-1007. Using an irreversible two-tissue compartment model, the area under curve (AUC) and the net influx rate Ki were calculated to assess the impact of different β-factors on the pharmacokinetic analysis of clinical PET brain data.ResultsIn general, RC and CVRC decreased with increasing β-factor in the phantom data. For small spheres (< 10 mm), and in particular for short acquisition times, low β-factors resulted in high variability and an overestimation of measured activity. Increasing the β-factor improves the variability, however at a cost of underestimating the measured activity. For the clinical data, AUC decreased and Ki increased with increased β-factor; a change in β-factor from 300 to 1000 resulted in a 25.5% increase in the Ki.ConclusionIn a complex dynamic dataset with variable acquisition times, the optimal β-factor provides a balance between accuracy and precision. Based on our results, we suggest a β-factor of 300–500 for quantitation of small structures with dynamic PET imaging, while large structures may benefit from higher β-factors.Trial registrationClinicaltrials.gov, NCT03951142. Registered 5 October 2019, https://clinicaltrials.gov/ct2/show/NCT03951142. EudraCT no 2018-003229-27. Registered 26 February 2019, https://www.clinicaltrialsregister.eu/ctr-search/trial/2018-003229-27/NO.

【 授权许可】

CC BY   
© Springer Nature Switzerland AG 2023

【 图 表 】

Fig. 2

Fig. 2

12864_2016_3440_Article_IEq8.gif

12936_2017_1963_Article_IEq63.gif

Fig. 2