Diagnostic Accuracy of an Abbreviated vs. a Full MRI Breast Protocol in Detecting Breast Cancer: An ROC Study

Background: /Objectives: Comparison of diagnostic accuracy of an abbreviated against a full protocol for breast MRI in detecting breast cancer. Diagnostic accuracy and cancer yield were compared using reporting and data system classification (BI-RADS) against a gold standard: biopsy confirmed carcinoma or negative follow-up on imaging. Diagnostic accuracy determined through receiver operator characteristic (ROC), and calculation of performance indicators. Methods: Thirty-five MRI breast examinations undertaken between January 2019 and December 2021: 20-biopsy confirmed carcinoma and 15-normal (negative follow-up on imaging), were retrospectively retrieved. Two radiologists independently reviewed the abbreviated protocol. Data collected compared with data from initial radiological MRI reports based on the full protocol. Lesions with BI-RADS 1 or 2 considered were considered negative while BI-RADS 3,4 or 5 positive. Results: Area under the curve obtained from each ROC curve (ROCAUC) was: 1 for the full protocol, 0.920 and 0.922 for Radiologist A and B respectively for the abbreviated protocol. Each protocol diagnosed the disease state (cancer yes/no) at a statistically significant level (p< 0.05). When using the abbreviated protocol, diagnostic accuracy was maintained. All cancers were detected by both radiologists on both abbreviated and full protocol (100% sensitivity). The specificity of the abbreviated protocol was significantly (p< 0.05) lower for both radiologists (A-73.3% and B -53.5%) compared to the full protocol (100% for both). Conclusions: Area under the curve obtained from each ROC curve (ROCAUC) was: 1 for the full protocol, 0.920 and 0.922 for Radiologist A and B respectively for the abbreviated protocol. Each protocol diagnosed the disease state (cancer yes/no) at a statistically significant level (p< 0.05). When using the abbreviated protocol, diagnostic accuracy was maintained. All cancers were detected by both radiologists on both abbreviated and full protocol (100% sensitivity). The specificity of the abbreviated protocol was significantly (p< 0.05) lower for both radiologists (A-73.3% and B -53.5%) compared to the full protocol (100% for both).