Computer-aided evaluation as an adjunct to revised BI-RADS Atlas: improvement in positive predictive value at screening breast MRI.

Abstract

To investigate whether kinetic features via magnetic resonance (MR)-computer-aided evaluation (CAE) can improve the positive predictive value (PPV) of morphological descriptors for suspicious lesions at screening breast MRI. One hundred and sixteen consecutive, suspiciously enhancing lesions detected at contralateral breast MRI screening in 116 women with newly-diagnosed breast cancers were included. Morphological descriptors according to the revised BI-RADS Atlas and kinetic features from MR-CAE were analysed. The PPV of each descriptor was analysed to identify subgroups in which PPV could be improved by the addition of MR-CAE. When biopsy recommendations were downgraded to follow-up in cases where there were both the absence of enhancement at a 50% threshold and the absence of delayed washout, PPV increased from 0.328 (95% CI, 0.249-0.417) to 0.500 (95% CI, 0.387- 0.613). Two ductal carcinoma in situ (DCIS) non-mass enhancement (NME) lesions were missed. Application of downgrading criteria to foci or masses led to increased PPV from 0.310 (95% CI, 0.216-0.419) to 0.437 (95% CI, 0.331-0.547) without missing cancers. MR-CAE has the potential to improve the PPV of breast MR imaging by reducing the number of false positives. When suspicious mass lesions do not show enhancement at a 50% threshold nor delayed washout, follow-up rather than biopsy can be considered. • MR-CAE has the potential to increase PPV at breast MRI screening. • Lesions without enhancement at 50% threshold and washout might be downgraded. • DCIS non-mass lesions might be false-negative cases at MR-CAE.