The Clinically Relevant Breast Imaging Audit

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Breast imaging practice audits necessitate standardized data definitions for comparable facility and staff performance evaluations.
• Key metrics rely on true positive (TP), true negative (TN), false positive (FP), and false negative (FN) data points.
• Defining 'true' for mammography involves confirming malignancy within one year of the examination.

Purpose of the Study:

• To identify reliable metrics for breast imaging practice audits.
• To address challenges in calculating sensitivity and specificity due to patient mobility and data availability.
• To highlight easily calculable internal metrics and the most accurate method for assessing interpreter skills.

Main Methods:

• Focus on internally calculable metrics: recall rate, positive predictive values (PPV1, PPV2, PPV3), and cancer detection rate.
• Discusses limitations of accuracy as a performance metric.
• Highlights the receiver operating characteristic (ROC) curve as the optimal method for evaluating interpreter performance.

Main Results:

• Standardized definitions are crucial for comparable breast imaging audits.
• Recall rate, PPV, and cancer detection rate are feasible internal metrics.
• Accuracy can yield misleading results; ROC curves provide a superior assessment of interpreter skill.

Conclusions:

• Breast imaging audits must use accepted definitions for meaningful comparisons.
• Internally calculable metrics like recall rate and cancer detection rate are valuable.
• ROC curves are the most accurate tool for assessing individual interpreter performance in breast imaging.