WE-E-217BCD-02: Update on the Upcoming ACR Digital Mammography QC Manual

Area of Science:

• Medical Imaging
• Image Registration
• Quantitative Analysis

Background:

• Evaluating image registration accuracy is crucial in medical imaging.
• Inter-patient variability can significantly impact registration quality.
• Existing methods may require ground truth or contoured structures.

Purpose of the Study:

• To introduce and define a novel metric, Distance Discordance (DD).
• To assess the utility of DD in quantifying registration accuracy.
• To demonstrate DD's applicability in evaluating inter-patient variability.

Main Methods:

• Developed a software phantom with variable geometric properties.
• Applied two B-Spline deformable image registration (DIR) algorithms (Elastix, Plastimatch).
• Calculated DD by measuring distances between corresponding anatomic points across deformed images and analyzed distributions (DDH).
• Validated the metric using Head & Neck patient data.

Main Results:

• Different DIR algorithms yielded distinct DD results.
• Elastix showed slightly lower mean DDH values (0-1.28 cm) than Plastimatch (0-1.43 cm).
• Head & Neck patient DDH data followed a lognormal distribution (mean 0.45 cm, std dev 0.42 cm).

Conclusions:

• Distance Discordance (DD) provides an interpretable, quantitative measure for registration goodness.
• DD can assess inter-patient variability's impact on registration across anatomy.
• The DD metric does not require ground truth or contoured structures, enhancing its applicability.