Influence of dominant sequence artifacts on generative AI for virtual contrast-enhancement in breast MRI
6 affiliations
- 1Institute of Radiology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Maximiliansplatz 3, 91054 Erlangen, Germany
- 2Department Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander- Universität Erlangen-Nürnberg, Erlangen, Germany
- 3Institute of Computer Science, Polish Academy of Science, Warsaw, Poland
- 4Department for Computer Science 5 (Pattern Recognition), Friedrich-Alexander-University Erlangen-Nuremberg (FAU), Martensstrasse 3, 91058 Erlangen, Germany
- 5Medical Informatics, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- 6Diagnostische Radiologie, Radiologie München, München, Germany
Abstract
Background: Virtual contrast-enhanced (vCE) breast MRI is a deep learning approach for generating contrast-like images from non-contrast acquisitions. As diffusion-weighted imaging (DWI) has emerged as a major contributor to enhancement synthesis, understanding the influence of DWI-related image artifacts is important for the robustness and potential clinical implementation of AI-based contrast-free imaging approaches. We evaluated the robustness of a pix2pix-based vCE synthesis model to artifact-contaminated DWI during training and inference.
Materials and methods: This IRB-approved, retrospective study included 1,974 multiparametric breast MRI examinations (mean age: 49 years) acquired with two clinical routine 3T MRI scanners (2017–2022). High b-value (b = 1500 s/mm2) DWI images were automatically classified for high-intensity (HI) and low-intensity (LI) artifacts using a pretrained neural network. A pix2pix 2D-GAN was trained either after excluding artifact-affected slices (vCE-clean) or using all available slices (vCE-all). Performance was evaluated using image similarity metrics, lesion-based analysis, reader assessment, and correlation of artifact scores with reconstruction errors.
Results: Excluding artifact-affected slices during training resulted in a small but statistically significant increase in whole-image structural similarity index (SSIM) with small effect size (vCE-all 0.843 ± 0.034; vCE-clean 0.849 ± 0.031, p < 0.001, dav = 0.20). However, lesion-based metrics and qualitative ratings of lesion enhancement and image quality were comparable between models. During Inference, strong LI DWI artifacts were associated with higher lesion-based errors.
Conclusion: Controlling for artifacts on diffusion weighted imaging (DWI) during vCE training only minorly influenced technical metrics but not perceived image quality/lesion enhancement. In contrast, severe signal-loss artifacts during inference impaired local performance, highlighting the importance of image quality control during potential deployment.
Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s12880-026-02996-8.
Related
- Generative enhancement of suboptimally opacified cerebral CT angiography for large-vessel occlusion detection: a multireader study
- Artifacts of orthodontic materials on cone beam computed tomography images.
- Reducing Stent-Induced Blooming Artifacts Using Virtual Monoenergetic Imaging Reconstructions in a Superficial Femoral Artery Phantom: An In-Vitro Study
- Pulmonary Resection for Bronchiectasis: Perioperative and Follow-Up Outcomes in a VATS-Dominant Cohort
- The Role of Dominant Lesion Site and SUVmax in the Depth of PSA Response in Metastatic Hormone-Sensitive Prostate Cancer
BibTeX
@article{brock2026influence,
title = {{Influence of dominant sequence artifacts on generative AI for virtual contrast-enhancement in breast MRI}},
author = {Luise Brock and Ameya Markale and Tri-Thien Nguyen and Marie Metz and Aju George and Shirin Heidarikahkesh and Hannes Schreiter and Julian Pfann Hossbach and Lorenz A. Kapsner and Sabine Ohlmeyer and Evelyn Wenkel and Michael Uder and Frederik B. Laun and Florian Knoll and Sebastian Bickelhaupt},
journal = {BMC Medical Imaging},
year = {2026},
month = nov,
doi = {10.1186/s12880-026-02996-8},
url = {https://doi.org/10.1186/s12880-026-02996-8},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Influence of dominant sequence artifacts on generative AI for virtual contrast-enhancement in breast MRI
AU - Luise Brock
AU - Ameya Markale
AU - Tri-Thien Nguyen
AU - Marie Metz
AU - Aju George
AU - Shirin Heidarikahkesh
AU - Hannes Schreiter
AU - Julian Pfann Hossbach
AU - Lorenz A. Kapsner
AU - Sabine Ohlmeyer
AU - Evelyn Wenkel
AU - Michael Uder
AU - Frederik B. Laun
AU - Florian Knoll
AU - Sebastian Bickelhaupt
PY - 2026
DA - 2026/11/03
JO - BMC Medical Imaging
DO - 10.1186/s12880-026-02996-8
UR - https://doi.org/10.1186/s12880-026-02996-8
AB - Background: Virtual contrast-enhanced (vCE) breast MRI is a deep learning approach for generating contrast-like images from non-contrast acquisitions. As diffusion-weighted imaging (DWI) has emerged as a major contributor to enhancement synthesis, understanding the influence of DWI-related image artifacts is important for the robustness and potential clinical implementation of AI-based contrast-free imaging approaches. We evaluated the robustness of a pix2pix-based vCE synthesis model to artifact-contaminated DWI during training and inference. Materials and methods: This IRB-approved, retrospective study included 1,974 multiparametric breast MRI examinations (mean age: 49 years) acquired with two clinical routine 3T MRI scanners (2017–2022). High b-value (b = 1500 s/mm2) DWI images were automatically classified for high-intensity (HI) and low-intensity (LI) artifacts using a pretrained neural network. A pix2pix 2D-GAN was trained either after excluding artifact-affected slices (vCE-clean) or using all available slices (vCE-all). Performance was evaluated using image similarity metrics, lesion-based analysis, reader assessment, and correlation of artifact scores with reconstruction errors. Results: Excluding artifact-affected slices during training resulted in a small but statistically significant increase in whole-image structural similarity index (SSIM) with small effect size (vCE-all 0.843 ± 0.034; vCE-clean 0.849 ± 0.031, p < 0.001, dav = 0.20). However, lesion-based metrics and qualitative ratings of lesion enhancement and image quality were comparable between models. During Inference, strong LI DWI artifacts were associated with higher lesion-based errors. Conclusion: Controlling for artifacts on diffusion weighted imaging (DWI) during vCE training only minorly influenced technical metrics but not perceived image quality/lesion enhancement. In contrast, severe signal-loss artifacts during inference impaired local performance, highlighting the importance of image quality control during potential deployment. Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s12880-026-02996-8.
SN - 1471-2342
ER -