Paired split-plot designs of multireader multicase studies.

Area of Science:

• Medical Imaging Analysis
• Statistical Study Design
• Radiology Research

Background:

• The fully crossed (FC) design is standard for multireader, multicase ROC studies comparing imaging modalities.
• FC designs require every reader to analyze every case for all modalities, increasing costs and reducing flexibility.

Purpose of the Study:

• To investigate paired split-plot (PSP) designs as a more cost-effective and flexible alternative to FC designs.
• To analytically compare the statistical precision and power of PSP and FC designs.

Main Methods:

• Utilized a [Formula: see text]-statistic based variance analysis for area under the ROC curve (AUC).
• Compared the statistical performance of PSP and FC designs under varying reader and case group configurations.
• Validated the iMRMC software for analyzing data from both FC and PSP designs.

Main Results:

• PSP designs can yield higher precision than FC designs with the same number of readers and readings.
• PSP designs can achieve equivalent statistical power to FC designs with fewer readings.
• PSP designs require a larger number of truth-verified patient cases, offering a trade-off between cost factors.

Conclusions:

• PSP designs present a statistically advantageous and flexible alternative for comparative imaging studies.
• The choice between FC and PSP designs allows for optimization based on specific cost and resource constraints.
• PSP designs were demonstrated effectively in a reader study comparing digital mammography and screen-film mammography.