Back to Search
Start Over
80. Advancements in PET image reconstruction algorithms: Preliminary results of their application
- Source :
- Physica Medica. 56:112
- Publication Year :
- 2018
- Publisher :
- Elsevier BV, 2018.
-
Abstract
- Purpose Important advancements have been made in PET image reconstruction algorithms, among the newest point spread function (PSF) and noise statistics modelling (NSM) can be accounted for. This work presents the preliminary results of the application of these reconstruction techniques on 3 state-of-the-art PET/CT scanners with and without time-of-flight (TOF) technique. Methods A Ge68 NEMA/IEC phantom (Data Spectrum, Durham, NC) filled with 68Ge in an epoxy resin was used. The activity concentrations in the spheres and in the background were measured with a radionuclide calibrator tested against a NIST traceable source and were respectively 40.67 kBq/ml and 10.58 kBq/ml at reference time (uncertainty ± 3)%. Nominal sphere-to-background ratio was 3.84. Total activity in the phantom was 108.4 MBq. The phantom was acquired for a prolonged time according to decay correction equivalent to 1.5 min per frame at the reference date on each PET scanner and reconstructed with clinical reconstruction parameters with traditional (OSEM) and by combining advanced modalities, when present: PSF, TOF, TOF + PSF, NSM. For each PET scanner and each reconstruction modality the maximum Recovery Coefficient (RCmax) values were evaluated for each visible sphere. RCmax were normalised to background activity to cancel calibration bias among scanners. Results The Figure presents the RCmax averaged on the 3 PET/CT scanners together with the RC limits suggested by EARL. Error bars represent one standard deviation. Conclusions the RC obtained with the 3 state-of-the-art non digital PET/CT scanners demonstrate clearly that, independently of the reconstruction method used, modern PET/CT scanner outperforms respect to older one described by EARL curve. Moreover, applying more advanced reconstruction techniques progressively enhances RC, NSM reconstruction being the most promising setup. Even if RCmax of different scanners converge at larger (>22 mm) and smaller (10 mm) spheres, the overestimations observed in the size between 1 and 2 cm are currently under investigation.
- Subjects :
- Point spread function
Scanner
business.industry
Computer science
Biophysics
General Physics and Astronomy
General Medicine
Standard deviation
Imaging phantom
Image reconstruction algorithm
Calibration
NIST
Radiology, Nuclear Medicine and imaging
Computer vision
Artificial intelligence
business
Decay correct
Subjects
Details
- ISSN :
- 11201797
- Volume :
- 56
- Database :
- OpenAIRE
- Journal :
- Physica Medica
- Accession number :
- edsair.doi...........f07d0dc2aa38eb7b49faceebbe5393a3
- Full Text :
- https://doi.org/10.1016/j.ejmp.2018.04.090