The Contrast Order: An Order-Based Image Quality Criterion for Nonlinear Beamformers

Area of Science:

• Medical Imaging
• Ultrasound Technology
• Image Quality Assessment

Background:

• Classical image quality metrics (contrast ratio, contrast-to-noise ratio) rely on summary statistics, which can be altered by modern ultrasound beamformers.
• Concerns exist that reported improvements in ultrasound image quality may stem from dynamic range changes rather than true information gain.
• Existing advanced metrics like generalized contrast-to-noise ratio (gCNR) have limitations due to reliance on noisy distribution estimates.

Purpose of the Study:

• Introduce novel, robust image quality criteria for evaluating modern ultrasound beamformers.
• Address limitations of classical and current advanced metrics for assessing lesion contrast.
• Develop criteria invariant to monotonic transformations and dynamic range variations.

Main Methods:

• Defined contrast order (CO) as the expected value of the sign of brightness difference between two regions of interest (ROIs).
• Developed a simple, unbiased estimator for CO with decreasing variance.
• Proposed effective contrast ratio (ECR) to calibrate CO with classical contrast ratio under specific statistics.

Main Results:

• CO is invariant under strictly monotonic transformations, depending only on relative ROI brightness ordering.
• CO's estimator variance decreases with increased samples per ROI.
• ECR aligns CO with classical contrast ratio under ideal Rayleigh-speckle statistics.

Conclusions:

• CO and ECR provide principled, order- and sign-preserving, dynamic-range-invariant criteria for lesion contrast evaluation.
• These new metrics offer a reliable alternative for assessing modern ultrasound beamformer performance.
• The proposed criteria enhance the accuracy of image quality assessment in ultrasound.