Although contrast-enhanced (CE) breast MRI is an important tool for cancer detection, intravenous gadolinium administration introduces additional considerations for patients and clinical workflows. AI based virtual contrast enhanced (VCE) MRI could help identify patients for whom acquired CE imaging may be avoidable without compromising diagnostic performance. We developed a framework that combines non contrast (NC) and VCE MRI predictions for cancer detection. On the held-out test set, cancer detection using NC MRI alone achieved an AUROC of 0.865. Combining the predictions from NC and VCE MRI improved AUROC to 0.901. Building on this complementary predictive value, we evaluated patient-level routing to progressively replace acquired CE examinations with the NC plus VCE pathway, prioritizing patients with the most confident NC predictions. This simulation suggests that diagnostic performance can be largely preserved while limiting acquired CE imaging to approximately 40% of patients. Together, these findings demonstrate the complementary value of VCE MRI and suggest its potential to support selective contrast use in breast cancer screening.
Figures & tables
Figure 1: Qualitative examples of virtual contrast enhanced MRI. Blue outlines indicate annotated lesions.
Exp.
Input / strategy
Held-out test set
Dataset-specific AUROC
AUROC
AUPRC
DUKE
MAMAMIA
ODELIA
1
NC MRI (NC Classifier)
0.865±0.007
0.831±0.020
0.861±0.005
0.897±0.017
0.795±0.012
2
VCE MRI (CE Classifier)
0.866±0.007
0.841±0.005
0.864±0.014
0.891±0.011
0.799±0.016
3
VCE MRI (VCE Classifier)
0.883±0.010
0.854±0.022
0.882±0.003
0.908±0.007
0.833±0.029
4
CE MRI (CE Classifier)
0.945±0.003
0.927±0.007
0.950±0.013
0.966±0.001
0.900±0.010
5
NC + VCE (NC + CE Classifiers)
0.901±0.005
0.874±0.005
0.888±0.008
0.939±0.012
0.841±0.003
Table 1: Cancer detection performance on the held-out test set, overall and by source dataset. Overall AUROC and AUPRC are reported as mean ± SD across three classifier repeats.
Figure 2: Selective contrast-sparing. AUROC and AUPRC versus the proportion of patients assigned to the NC plus VCE pathway. Shaded regions indicate 95% randomization intervals under random patient assignment.
Center for Big Data and Intelligent Medicine, The First Affiliated Hospital of Chongqing Medical University, Chongqing, China · Key Laboratory of Digital Health and Intelligent Medicine, Chongqing Municipal Health Commission, Chongqing, China · Chongqing Translational Medicine Center, Chongqing, China
German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Germany · Medical Faculty, Heidelberg University, Germany · Faculty of Mathematics and Computer Science, Heidelberg University, Germany +3
Computer Sciences Department, University of Oviedo, Asturias, Spain · Biomedical Engineering Center (BME), University of Oviedo, Asturias, Spain · Electrical Engineering Department, University of Oviedo, Asturias, Spain +6