Ultrafast versus Standard DCE-MRI in Breast Lesion and Background Parenchymal Enhancement Differentiation
2Meirav Center, Sheba Medical Center, Israel
3Gray Faculty of Medical & Health Sciences, Tel Aviv University, Israel
Background: Background parenchymal enhancement (BPE) on breast MRI can mimic true lesions, contributing to false-positive interpretations and unnecessary biopsies. Ultrafast DCE (UF-DCE) MRI provides substantially higher temporal resolution than standard DCE-MRI, but whether it can substitute for or complement standard DCE-MRI remains unclear. This study compared the two acquisitions for two sequential tasks: distinguishing true lesions from BPE, and distinguishing malignant from benign lesions.
Methods: Breast MRI examinations from 276 patients, comprising 344 pathology-proven lesions (176 malignant, 168 benign) and 892 BPE foci matched in size and shape to the lesions, were retrospectively analyzed. Identical radiomic features were extracted from UF-DCE (10 timepoints) and standard DCE (4 phases). Two binary tasks were evaluated separately, detection (lesion versus BPE) and characterization (malignant versus benign), using five model families with nested 5-fold cross-validation grouped by examination, with feature selection and correlation-based pruning restricted to training folds. Performance was compared statistically between acquisitions.
Results: For distinguishing true lesions from BPE, standard DCE outperformed UF-DCE (AUC, 0.979 vs. 0.941; p<0.001), reaching 93.6% sensitivity and 90.2% specificity, reflecting greater homogeneity in BPE relative to true lesions. In contrast, for distinguishing malignant from benign lesions, UF-DCE outperformed standard DCE (AUC, 0.827 vs. 0.787; p=0.023), reaching 72.7% sensitivity and 72.6% specificity, an advantage concentrated in texture features from the later ultrafast timepoints and confirmed, under a matched feature-count control, not to reflect UF-DCE`s larger feature pool. Both directions were consistent across all five model families tested, and combining both acquisitions improved neither task.
Conclusions: UF-DCE and standard DCE are complementary rather than interchangeable. Standard DCE is better suited to exclude BPE-related false-positives, while UF-DCE better distinguishes malignant from benign lesions. These findings support a sequence-specific interpretation strategy, in which standard DCE confirms that a finding is a true lesion and UF-DCE subsequently characterizes it.