Wednesday, November 29 | 9:40 a.m.-9:50 a.m. | W3-SSBR08-2 | Room S406A
A deep learning-based technique could pave the way for lower doses of gadolinium-based contrast agents (GBCAs) in breast MRI, according to this presentation.
GBCAs are widely used in breast cancer screening and evaluating treatment response, but various safety concerns have resulted in a need for dose reduction, according to Srivathsa Pasumarthi Venkata of AI software developer Subtle Medical and colleagues.
In this study, the researchers repurposed a deep learning-based dose reduction model that was trained to synthesize full-dose contrast-enhanced brain MRI images from precontrast and 10% low-dose images. Now, it was used to synthesize post-contrast breast MRI images from precontrast images and synthetic low-dose breast images, which were simulated by a vision transformer using pre- and postcontrast images.
After quantitatively and qualitatively assessing the algorithm’s performance on a dataset of dynamic contrast-enhanced (DCE) breast MRI data from 60 patients, the researchers concluded that their method was feasible.
“Enhancement patterns of [deep learning]-synthesized DCE images are similar to standard-dose images and can be used for contrast dose reduction in breast MRI exams, reducing both patient and environmental exposure to GBCAs,” the authors wrote.
They also noted that their deep-learning model could be extended for use on real low-dose images.
How did researchers achieve these results? Check out this Wednesday morning talk to learn more.
