Medical AI / 2025
Virtual Contrast Enhancement in Breast MRI
Deep learning workflow for contrast-agent-free breast MRI using GAN-based virtual contrast enhancement.
Problem
Contrast-enhanced breast MRI is clinically valuable, but contrast agents add cost, time, and patient burden. The research question is whether deep learning can generate virtual contrast images that are useful enough to support contrast-agent-free workflows.
Approach
The project uses generative adversarial networks for virtual contrast enhancement in breast MRI, with emphasis on comparing generated images against real contrast images and understanding where the model is reliable.
Architecture
Diagram placeholder: breast MRI input, preprocessing, GAN model, virtual contrast output, comparison to real contrast images, and downstream research evaluation.
Results
The work connects model development with evaluation questions around similarity to real contrast images, usability, and reproducibility in a research setting.
Learnings
Medical generative AI needs strong evaluation and careful workflow design. The model output is only useful when the surrounding data handling, testing, and clinical comparison are reliable.