Virtual DCE breast MRI shows similar BI-RADS accuracy but lower sensitivity than gadolinium-enhanced DCE

Deep-learning virtual dynamic contrast-enhanced (DCE) breast MRI reached 92% sensitivity vs 96% for standard gadolinium-enhanced DCE, with accuracy 76% vs 77% and no significant differences in BI-RADS or background enhancement in a 592-exam study.

• The virtual DCE model was a 2D U-Net generating contrast-like series from unenhanced T1-weighted, T2-weighted, and diffusion-weighted acquisitions.
• Mean paired lesion-size difference (DCE minus virtual DCE) was -2.2 mm in the holdout test set and 0.5 mm in a secondary test set.
• Background parenchymal enhancement and artifact ratings did not differ significantly (p=1.0); one reader's confidence was lower for virtual DCE (p=0.072).

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.