90Y ‐PET imaging: Exploring limitations and accuracy under conditions of low counts and high random fraction. Issue 7 (22nd June 2015)

• Record Type: Journal Article
• Authors: Carlier, Thomas; Willowson, Kathy P.; Fourkal, Eugene; Bailey, Dale L.; Doss, Mohan; Conti, Maurizio
• Is Part Of: Medical physics. Volume 42:Issue 7(2015)
• Page Start: 4295
• Page End: 4309
• Publication Date: 2015-06-22
• Subjects: expectation‐maximisation algorithm -- image denoising -- image reconstruction -- iterative methods -- liver -- medical image processing -- optical transfer function -- phantoms -- positron emission tomography -- smoothing methods -- yttrium; Positron emission tomography (PET) -- Noise -- Reconstruction -- Numerical approximation and analysis -- Probability theory, stochastic processes, and statistics; Biological material, e.g. blood, urine; Haemocytometers -- Digital computing or data processing equipment or methods, specially adapted for specific applications -- Image data processing or generation, in general -- Image enhancement or restoration, e.g. from bit‐mapped to bit‐mapped creating a similar image -- Scintigraphy -- Measuring half‐life of a radioactive substance; 90Y ‐PET -- radioembolization -- quantification; Positron emission tomography -- Computed tomography -- Medical image noise -- Image reconstruction -- Medical image reconstruction -- Liver -- Image scanners -- Photons; Medical physics -- Periodicals; Medical physics; Geneeskunde; Natuurkunde; Toepassingen; Biophysics; Periodicals; Periodicals; Electronic journals; 610.153
• DOI: 10.1118/1.4922685
• Languages: English
• ISSNs: 0094-2405

Abstract

Abstract : Purpose: 90 Y ‐positron emission tomography (PET) imaging is becoming a recognized modality for postinfusion quantitative assessment following radioembolization therapy. However, the extremely low counts and high random fraction associated with 90 Y ‐PET may significantly impair both qualitative and quantitative results. The aim of this work was to study image quality and noise level in relation to the quantification and bias performance of two types of Siemens PET scanners when imaging 90 Y and to compare experimental results with clinical data from two types of commercially available 90 Y microspheres. Methods: Data were acquired on both Siemens Biograph TruePoint [non‐time‐of‐flight (TOF)] and Biograph microcomputed tomography (mCT) (TOF) PET/CT scanners. The study was conducted in three phases. The first aimed to assess quantification and bias for different reconstruction methods according to random fraction and number of true counts in the scan. The NEMA 1994 PET phantom was filled with water with one cylindrical insert left empty (air) and the other filled with a solution of 90 Y . The phantom was scanned for 60 min in the PET/CT scanner every one or two days. The second phase used the NEMA 2001 PET phantom to derive noise and image quality metrics. The spheres and the background were filled with a 90 Y solution in an 8:1 contrast ratio and four 30 min acquisitions were performed over a one week period. Finally, 32 patient data (8 treated with Therasphere ® and 24 with SIR‐Spheres ® ) were retrospectively reconstructed and activity in the whole field of view and the liver was compared to theoretical injected activity. Results: The contribution of both bremsstrahlung and LSO trues was found to be negligible, allowing data to be decay corrected to obtain correct quantification. In general, the recovered activity for all reconstruction methods was stable over the range studied, with a small bias appearing at extremely high random fraction and low counts for iterative algorithms. Point spread function (PSF) correction and TOF reconstruction in general reduce background variability and noise and increase recovered concentration. Results for patient data indicated a good correlation between the expected and PET reconstructed activities. A linear relationship between the expected and the measured activities in the organ of interest was observed for all reconstruction method used: a linearity coefficient of 0.89 ± 0.05 for the Biograph mCT and 0.81 ± 0.05 for the Biograph TruePoint. Conclusions: Due to the low counts and high random fraction, accurate image quantification of 90 Y during selective internal radionuclide therapy is affected by random coincidence estimation, scatter correction, and any positivity constraint of the algorithm. Nevertheless, phantom and patient studies showed that the impact of number of true and random coincidences on quantitative results was found to be limited as long as ordinary Poisson ordered subsets expectation maximization reconstruction algorithms with random smoothing are used. Adding PSF correction and TOF information to the reconstruction greatly improves the image quality in terms of bias, variability, noise reduction, and detectability. On the patient studies, the total activity in the field of view is in general accurately measured by Biograph mCT and slightly overestimated by the Biograph TruePoint.