Clinical evaluation of a block sequential regularized expectation maximization reconstruction algorithm in 18F-FDG PET/CT studies

Authors:
  • PSPaul Stolzmann
    Paul Stolzmann
  • SWScott D. Wollenweber
    Scott D. Wollenweber
  • MHMartin Hüllner
    Martin Hüllner
  • YHYahya A. Hakami
    Yahya A. Hakami
  • FBFelipe de Galiza Barbosa
    Felipe de Galiza Barbosa
  • GSGustav K. von Schulthess
    Gustav K. von Schulthess
  • CPCarsten Pietsch
    Carsten Pietsch
  • PVPatrick Veit-Haibach
    Patrick Veit-Haibach
Publisher website
Google Scholar
Abstract
To investigate the clinical performance of a block sequential regularized expectation maximization (BSREM) penalized likelihood reconstruction algorithm in oncologic PET/computed tomography (CT) studies. A total of 410 reconstructions of 41 fluorine-18 fluorodeoxyglucose-PET/CT studies of 41 patients with a total of 2010 lesions were analyzed by two experienced nuclear medicine physicians. Images were reconstructed with BSREM (with four different β values) or ordered subset expectation maximization (OSEM) algorithm with/without time-of-flight (TOF/non-TOF) corrections. OSEM reconstruction postfiltering was 4.0 mm full-width at half-maximum; BSREM did not use postfiltering. Evaluation of general image quality was performed with a five-point scale using maximum intensity projections. Artifacts (category 1), image sharpness (category 2), noise (category 3), and lesion detectability (category 4) were analyzed using a four-point scale. Size and maximum standardized uptake value (SUVmax) of lesions were measured by a third reader not involved in the image evaluation. BSREM-TOF reconstructions showed the best results in all categories, independent of different body compartments. In all categories, BSREM non-TOF reconstructions were significantly better than OSEM non-TOF reconstructions (P<0.001). In almost all categories, BSREM non-TOF reconstruction was comparable to or better than the OSEM-TOF algorithm (P<0.001 for general image quality, image sharpness, noise, and P=1.0 for artifact). Only in lesion detectability was OSEM-TOF significantly better than BSREM non-TOF (P<0.001). Both BSREM-TOF and BSREM non-TOF showed a decreasing SUVmax with increasing β values (P<0.001) and TOF reconstructions showed a significantly higher SUVmax than non-TOF reconstructions (P<0.001). The BSREM reconstruction algorithm showed a relevant improvement compared with OSEM reconstruction in PET/CT studies in all evaluated categories. BSREM might be used in clinical routine in conjunction with TOF to achieve better/higher image quality and lesion detectability or in PET/CT-systems without TOF-capability for enhancement of overall image quality as well.

This publication has 29 references indexed in Scilit: