Nuclear Medicine Uniformity Assessment Using 2D Noise Power Spectrum
J Nelson1*, O Christianson2, B Harkness3, M Madsen4, E Mah5, S Thomas6, H Zaidi7, E Samei8, (1) Duke University Medical Center, Durham, NC, (2) Duke University Medical Center,Durham, NC, (3) Henry Ford Hospital System, Detroit, MI, (4) University of Iowa, Iowa City, IA, (5) Medical University of South Carolina, Charleston, SC, (6) University of Cincinnati Medical Center, Cincinnati, OH, (7) Geneva University Hospital, Geneva, Switzerland(8) Duke University Medical Center, Durham, NCSU-D-217A-3 Sunday 2:15:00 PM - 3:00:00 PM Room: 217A
Nuclear medicine quality control programs require daily evaluation for the presence of potential non-uniformities by commonly utilizing a traditional pixel value-based assessment (Integral CFOV Uniformity). While this method effectively captures regional non-uniformities in the image, it does not adequately reflect subtle periodic structures that are visually apparent and clinically unacceptable, therefore requiring the need for additional visual inspection of the image. The goal of this project was to develop a new uniformity assessment metric by targeting structural patterns and more closely correlating with visual inspection.
The new quantitative uniformity assessment metric is based on the 2D Noise Power Spectrum (NPS). A full 2D NPS was performed on each image. The NPS was thresholded to remove quantum noise and further filtered by the visual response function. A score, the Structure Noise Index (SNI), was then applied to each based on the average magnitude of the structured noise in the processed image.
To verify the validity of the new metric, 50 daily uniformity images with varying degrees of visual structured and non-structured non-uniformity were scored by 5 expert nuclear medicine physicists. The correlation between the visual score and SNI were assessed. The Integral CFOV was also compared against the visual score.
Our new SNI assessment metric compared to the Integral CFOV showed in increase in sensitivity from 67% to 100% in correctly identifying structured non-uniformities. The overall positive predictive value also increased from 55% to 72%.
Our new uniformity metric correlates much more closely with visual assessment of structured non-uniform NM images than the traditional pixel-based method. Using this new metric in conjunction with the traditional pixel value-based assessment will allow a more accurate quantitative assessment of nuclear medicine uniformity.