Program Information
PET Image Noise Variability and Its Consequences for Quantifying Tumor Hypoxia
R Kueng1,2, P Manser1 , M K Fix1 , B Driscoll2 , M F M Stampanoni3 , H Keller*2 , (1) Division of Medical Radiation Physics and Department of Radiation Oncology, Inselspital, Bern University Hospital, and University of Bern, Switzerland, (2) Radiation Medicine Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada, (3) Institute for Biomedical Engineering, University of Zurich and Swiss Federal Institute of Technology (ETH)
Presentations
SU-G-IeP4-13 (Sunday, July 31, 2016) 5:30 PM - 6:00 PM Room: ePoster Theater
Purpose:
The values in a PET image which represent activity concentrations of a radioactive tracer are influenced by a large number of parameters including patient conditions as well as image acquisition and reconstruction. This work investigates noise characteristics in PET images for various image acquisition and image reconstruction parameters.
Methods:
Different phantoms with homogeneous activity distributions were scanned using several acquisition parameters and reconstructed with numerous sets of reconstruction parameters. Images from six PET scanners from different vendors were analyzed and compared with respect to quantitative noise characteristics. Local noise metrics, which give rise to a threshold value defining the metric of hypoxic fraction, as well as global noise measures in terms of noise power spectra (NPS) were computed. In addition to variability due to different reconstruction parameters, spatial variability of activity distribution and its noise metrics were investigated. Patient data from clinical trials were mapped onto phantom scans to explore the impact of the scanner’s intrinsic noise variability on quantitative clinical analysis.
Results:
Local noise metrics showed substantial variability up to an order of magnitude for different reconstruction parameters. Investigations of corresponding NPS revealed reconstruction dependent structural noise characteristics. For the acquisition parameters, noise metrics were guided by Poisson statistics. Large spatial non-uniformity of the noise was observed in both axial and radial direction of a PET image. In addition, activity concentrations in PET images of homogeneous phantom scans showed intriguing spatial fluctuations for most scanners. The clinical metric of the hypoxic fraction was shown to be considerably influenced by the PET scanner’s spatial noise characteristics.
Conclusion:
We showed that a hypoxic fraction metric based on noise characteristics requires careful consideration of the various dependencies in order to justify its quantitative validity. This work may result in recommendations for harmonizing QA of PET imaging for multi-institutional clinical trials.
Contact Email: