Enhanced FDG-PET tumor imaging with correlation-coefficient filtered influx-constant images

Kenneth R. Zasadny, Richard L. Wahl

Research output: Contribution to journalArticlepeer-review

47 Scopus citations


Although FDG is an excellent PET tumor imaging agent, residual tracer activity in normal structures such as blood vessels and the liver can impair the detection of small or modestly tracer-avid tumors. Since tumors generally have a continuous influx of FDG over time while normal tissues generally show tracer efflux, we produced and optimized correlation-coefficient constrained influx-constant 'parametric' images to maximize tumor visualization but minimize background and artifacts for FDG-PET cancer imaging. Methods: Influx-constant image sets were generated in 17 patients with various cancers for a range of correlation-coefficient constraint values. Quantitative evaluation of the parametric PET images was performed. Results: Image noise was reduced 70% (mean) with no loss of tumor signal for r ≥ 0.90 constraint versus no constraint. Higher (0.95-0.99) constraints improved tumor-to- normal ratios but resulted in some loss of tumor signal. Mild constraints (0- 0.85) produced more background artifacts than higher constraints, though all correlation constraints improved tumor-to-normal ratios over the single 50- 60-min acquisition frame. Conclusion: Correlation-coefficient filtered parametric imaging, especially with an r-value constraint of ≥0.90, enhances tumor-to-normal contrast for FDG-PET and appears promising for improving lesion detectability.

Original languageEnglish (US)
Pages (from-to)371-374
Number of pages4
JournalJournal of Nuclear Medicine
Issue number2
StatePublished - Feb 1 1996


  • PET
  • cancer
  • emission computed tomography
  • flurodeoxyglucose
  • tumor imaging

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging


Dive into the research topics of 'Enhanced FDG-PET tumor imaging with correlation-coefficient filtered influx-constant images'. Together they form a unique fingerprint.

Cite this