Multi-reader multi-case analysis of variance software for diagnostic performance comparison of imaging modalities

Area of Science:

• Medical Imaging Analysis
• Statistical Software Development
• Diagnostic Performance Evaluation

Background:

• Comparing diagnostic imaging modalities requires specialized statistical methods due to overlapping readers and cases.
• Existing analytical methods for multi-reader multi-case (MRMC) studies can be complex.
• Accurate statistical comparisons are crucial for evaluating and selecting optimal imaging techniques.

Purpose of the Study:

• To introduce MRMCaov, a novel R software package designed for MRMC comparisons of diagnostic imaging modalities.
• To provide an open-source tool for statistically comparing the performance of multiple imaging modalities.
• To facilitate the analysis, plotting, and presentation of results from MRMC studies.

Main Methods:

• Utilizes analysis of variance (ANOVA) methods based on Obuchowski and Rockette (1995) and Hillis et al. (2005, 2007, 2008, 2018).
• Supports parametric and non-parametric estimation of receiver operating characteristic (ROC) curves.
• Incorporates various inference options for readers and cases, and covariance estimation techniques (DeLong, jackknife, unbiased).

Main Results:

• MRMCaov enables comparison of reader performance metrics, such as area under the ROC curve (ROC AUC).
• The package provides modality-specific performance estimates with confidence intervals and p-values.
• It offers flexibility in study designs (factorial, nested, partially paired) and inference frameworks.

Conclusions:

• MRMCaov is a comprehensive, open-source R package for advanced MRMC statistical analysis in medical imaging.
• The software enhances the ability to accurately compare diagnostic performance across multiple imaging modalities.
• It offers a user-friendly interface for complex statistical analyses, plotting, and result presentation across different operating systems.