Combining a Breast Apparent Diffusion Coefficient (ADC-B) Category System with BI-RADS Assessment Improves Specificity of Breast Lesions Diagnosis

Abstract Purpose To investigate the diagnostic performance of the predefined ADC-B category system in differentiating malignant from benign breast lesions and to compare with the BI-RADS. Methods This was a single-institution retrospective study of patients who underwent breast MRI between April 2019 and May 2023. Dynamic contrast enhanced (DCE) MRI and diffusion weighted imaging (DWI) were performed using a 3-T MRI system. Data on lesion morphology (mass, non-mass), size and ADC were collected. Histology was the standard of reference. The analysis assessed the inter-reader agreement in measuring ADC and ADC-B category using the intraclass correlation coefficient (ICC), as well as the diagnostic performance based on the receiver operating characteristic (ROC) curve. Results A total of 376 lesions in 358 women (mean age, 46.29 years; SD, 11.03) with pathologic results (236 malignant and 140 benign) were included. The inter-reader agreement was excellent in measuring ADC (ICC = 0.991) and assessing ADC-B category (ICC = 0.967). Overall diagnostic performance for ADC-B category (AUC, 0.858; 95% CI: 0.816-0.894) was higher than for BI-RADS (AUC, 0.805; 95% CI: 0.759-0.846; P = 0.029). The AUC of ADC-B category combined with BI-RADS reached 0.870 (95% CI: 0.829-0.904) for ADC-measurable lesions and 0.861 (95% CI: 0.822-0.894) for all lesions. The diagnostic combination significantly improves the specificity of BI-RADS (from 17.1% to 49.5% for ADC measurable lesions and from 20% to 45.6% for all lesions; P < 0.001) while maintaining sensitivity. Conclusion The combination of predefined ADC-B category and BI-RADS has the potential to enable classification of breast lesion types with high accuracy. Advance in knowledge While DWI has been incorporated into clinical MRI protocols at numerous medical centers, it has not been included in the official BI-RADS criteria. Adding ADC-B category system to BI-RADS classification significantly improves the specificity of breast lesion classification without decreasing sensitivity compared to the BI-RADS alone.