Clinical Significance of Bayesian penalized-likelihood and Motion Correction PET Reconstruction Algorithms


Kaya G., Pala H., Tuncel M., Uğur Ö.

Annual Congress of the European Association of Nuclear Medicine October 15-19, 2022 Barcelona, Spain, Barcelona, Spain, 15 - 19 October 2022, pp.163

  • Publication Type: Conference Paper / Summary Text
  • Doi Number: 10.1007/s00259-022-05924-4
  • City: Barcelona
  • Country: Spain
  • Page Numbers: pp.163
  • Hacettepe University Affiliated: Yes

Abstract

ayesian penalized-likelihood (BPL) iterative positron emission tomography (PET) reconstruction algorithms were introduced by many vendors. Decreasing noise and improving lesion detection are the claimed benefits of these algorithms. Motion-correction algorithms were also proposed to enhance image quality by correcting respiratory motions. This study aims to assess the clinical contribution of the BPL and motion-correction algorithms and evaluate their effects on the quantification of SUVmax. Materials and Methods: This study evaluated PET images of patients suffering from FDG avid malignant disease. PET images were reconstructed by OSEM, BPL and motion-corrected BPL algorithms. Two experienced nuclear medicine physicians assessed the clinical contributions of additional findings obtained by BPL algorithms. Visual improvements in lesion conspicuity and detection of additional malignant lesions were noted. SUVmax measurements of the liver, lesions under 10 mm, and lesions larger than 25 mm by OSEM, BPL, and motion-corrected BPL algorithms were also obtained and compared retrospectively. Results: The study included 199 patients (F/M: 103/96, median age:53 (range 5-87)) with a diagnosis of the breast (50), lymphoproliferative (45), lung (26), colorectal (12), head&neck (11), hepatobiliary-pancreatic

(8), gastric (8), malign melanoma (6), sarcoma (9) and other (24) malignancies. 597 PET images (OSEM, BPL and motion-corrected- BPL) of 199 PET studies were analyzed. BPL reconstruction provided an increase in lesion conspicuity, for 3 patients with 3 liver lesions (dominant; 12 mm, OSEM-SUVmax: 2.9, BPL-SUVmax: 4.6), 1 lung, 1 adrenal, 1 gastric, 1 breast lesion and 1 axillary lymph node at different patients for a total 8 patients (median lesion size: 9 mm). In only one patient an additional lesion was detected in the liver that was not visible in OSEM reconstructed images. There was no significant change between BPL methods with or without motion correction. No significant additional finding was observed that affected clinical follow-up. The liver’s mean SUVmax values were not statistically significant (2.5, 2.6 and 2.7 for OSEM, BPL and motion- corrected-BPL images respectively, p:0.67). Median SUVmax of lesions for under-10mm group is 3.26 (range 1-9) for OSEM and 4.50 (range 1-14) for BPL algorithms (increased 38% p<0.001) and also for larger-than-25mm group is 13.41 (range 1-35) for OSEM and 14.67 (range 2-43) for BPL algorithms (increased 9% p<0.001). Conclusion: This study showed that SUVmax value changes of small lesions measured significantly higher than larger lesions via BPL algorithm compared to OSEM. For the follow-up smaller tumors SUVmax values from the same reconstruction algorithm are crucial.