Improved nuclear medicine uniformity assessment with noise texture analysis

Area of Science:

• Nuclear Medicine Imaging
• Medical Physics
• Image Analysis

Background:

• Gamma cameras require routine uniformity performance evaluation due to environmental and system vulnerabilities.
• Current pixel value-based metrics often fail to detect subtle periodic structures, necessitating subjective visual inspections.
• Accurate identification of nonuniformities is critical for reliable nuclear medicine imaging.

Purpose of the Study:

• To develop, test, and validate a novel uniformity analysis metric for nuclear medicine flood-field images.
• To create a quantitative metric capable of identifying subtle structures and patterns missed by traditional methods.
• To improve the accuracy and objectivity of uniformity performance evaluations in gamma cameras.

Main Methods:

• Introduced the Structured Noise Index (SNI), based on the 2D noise power spectrum (NPS).
• Quantum noise contribution was subtracted from the NPS to isolate image artifacts.
• An observer study with nuclear medicine physicists validated the SNI against visual assessment and pixel-based metrics.

Main Results:

• The SNI demonstrated a strong correlation with expert visual scores (ρ = 0.86), outperforming traditional metrics (ρ = 0.59-0.58).
• SNI achieved 100% sensitivity in detecting both structured and non-structured nonuniformities, compared to 54-62% for others.
• SNI showed a higher positive predictive value (87%) than integral UFOV (67%) and CFOV (50%) metrics.

Conclusions:

• The SNI accurately identifies and quantifies flood-field nonuniformities, closely correlating with expert visual perception.
• SNI offers superior performance over traditional pixel value-based analyses for detecting visually apparent nonuniformities.
• This novel metric can enhance the objectivity of uniformity assessments and potentially reduce reliance on subjective visual inspections.