PT  - JOURNAL ARTICLE
AU  - Paul Kinahan
AU  - Adam Opanowski
AU  - Matthew Miller
TI  - Impact of PSF reconstruction on clinical trial qualification and patient SUVs
DP  - 2017 May 01
TA  - Journal of Nuclear Medicine
PG  - 717--717
VI  - 58
IP  - supplement 1
4099  - http://jnm.snmjournals.org/content/58/supplement_1/717.short
4100  - http://jnm.snmjournals.org/content/58/supplement_1/717.full
SO  - J Nucl Med2017 May 01; 58
AB  - 717Objectives: Qualification for clinical trials using PET/CT scans defines the technical and behavioral performance levels and quality control specifications for single- and multi-center clinical trials. While the emphasis is on clinical trials, this process is also intended to apply for clinical practice where quantitation is used. The procedures for scanner qualification and patient imaging in many trials are silent on the use of point-spread-function (PSF) image reconstruction methods. This is despite the well-known ringing artifact that can appear in PET images when PSF methods are used. Our goal was to understand the prevalence and potential impact of PSF methods in a multi-center clinical trial using PET imaging.Methods: We used scanner qualification results and patient imaging parameters from an ongoing multi-center clinical trial using PET imaging. The imaging procedures explicitly require that PSF methods should not be used for scanner qualification results or patient imaging. We reviewed the phantom and patient images submitted by the 17 qualified sites. Assessment of the results used quantitative and qualitative assessments.Results: Approximately 25% of the sites submitted phantom and/or patient images using PSF methods. Inspection of the phantom qualification images clearly demonstrated the ringing effect at edges. However, quantitative assessment of the phantom data using standard metrics (SUVmax, SUVmean, and standard deviation within the region of interest) used for site qualification was not able to distinguish between images reconstructed with or without PSF methods. Conversely, patient images showed anecdotal evidence of increases of approximately 30% in all standard metrics for small lesions.Conclusion: The use of PSF-based image reconstruction methods appears to be increasingly prevalent in clinical practice and may affect studies even if the trial requests results be submitted without PSF-based reconstruction. Whether or not PSF methods are used is not apparent from patient images or standard metrics used for site qualification - only careful examination of metadata from the DICOM header was helpful in this regard. Given that EORTC and PERCIST criteria use change thresholds of ±25% and ±30%, the use of PSF methods which can cause increases of 30% or higher for the same image can be a confounding variable in clinical trials using multiple sites. More effort is needed to understand the impact of PSF-based image reconstruction methods on the results of clinical trials. Research Support: None.