DeepVEST: Deep Learning-based Vessel Segmentation and Erasure on Breast MR Images for Improved Lesion Assessment

Purpose To investigate whether selective removal of vascular structures can improve lesion visibility and interpretability in maximum intensity projection (MIP) images derived from dynamic contrast-enhanced MRI. Materials and Methods A retrospective analysis was conducted using breast MRI scans from the Duke-Breast-Cancer-MRI dataset for model development. Deep Learning-based Vessel sEgmentation and eraSure in Breast MRI (DeepVEST), a deep learning method for automatic vessel segmentation and removal, was developed. A reader study with five breast radiologists was conducted to evaluate the effect of vessel-removed MIPs on lesion assessment, using images from the Duke-Breast-Cancer-MRI and Advanced-MRI-Breast-Lesions datasets. Segmentation performance was evaluated against manually annotated vessel segmentations using the Dice similarity coefficient, and perceived usefulness and quality were measured through reader study metrics. Interreader agreement on vessel removal quality and artifact impact was calculated using the Gwet agreement coefficient. Results DeepVEST achieved a Dice similarity coefficient of 0.611 for vessel segmentation. In the reader study (150 assessments, 600 responses), vessels partially or fully obscured lesion margins in 91 of 150 assessments (60.7%). Vessel removal effectiveness was rated 3.820 ± 0.749 on a 5-point Likert scale. Artifacts were reported in 45 of 150 assessments (30.0%), with a low mean severity score of 0.493 ± 0.900 (scale of 0-5), indicating minimal effect of image quality. Substantial interreader agreement was observed for vessel removal quality (Gwet agreement coefficient of 0.703) and artifact evaluation (Gwet agreement coefficient of 0.736). Conclusion The DeepVEST method enabled automatic vessel segmentation and removal on breast MR MIP images, showing high effectiveness with minimal artifacts. Most readers found it helpful for lesion assessment in challenging cases. Keywords: Segmentation, Breast, Neural Networks, Image Postprocessing, MR-Imaging, Convolutional Neural Network (CNN), Breast MRI, Deep Learning, Vessel Segmentation, Vessel Removal, Reader Study Supplemental material is available for this article. © RSNA, 2026 See also the commentary by Tayebi Arasteh in this issue.