Scientists at the Center for Advanced Imaging Innovation and Research are sharing a 3D vision transformer-based neural network for fast and accurate reconstruction of electrical properties from MRI data—a new resource for investigations on magnetic resonance electrical properties tomography, or MR-EPT.
MR-EPT is used to estimate electric conductivity and relative permittivity in tissues based on MRI data. This kind of analysis provides valuable information about interactions between electromagnetic waves and biological tissues, and has the potential to enable many promising applications in biomedical research and medicine. However, MR-EPT is computationally challenging and traditional methods tend to be slow and imprecise.
NYU Langone’s imaging scientists are now making available a deep learning model that outperforms traditional methods.
Ilias Giannakopoulos, PhD, postdoctoral fellow at NYU Langone Health and lead author of this research, was honored for this advance at the 2024 meeting of the International Society for Magnetic Resonance in Medicine (ISMRM) in Singapore with a magna cum laude award, and recently presented this work at the 2024 joint workshop on MR phase, magnetic susceptibility and electrical properties mapping in Chile (EMTP Chile) in a session dedicated to the state of the art in EPT.
In a peer-reviewed report published in the journal Magnetic Resonance in Medicine, the Dr. Giannakopoulos and coauthors describe their neural network as “an important step towards the development of clinically-usable in vivo [electrical properties] reconstruction protocols.”
A version of this post first appeared on the CAI2R LinkedIn.
Related Resource
A neural network for reconstructing electrical properties of materials and biological tissue from MRI data, plus synthetic phantoms.
Related Story
Ilias Giannakopoulos, postdoctoral fellow in MRI, talks about how electromagnetic waves interact with the body, why matrix compression matters, and where he finds inspiration.