In this project we pursue a three-pronged research strategy to obtain accurate maps of the electrical properties of biological tissue.
The electrical properties (EP) of biological tissue, namely electric conductivity and relative permittivity, provide quantitative information that has the potential to advance technical, medical, and scientific applications of MRI. These include: more precise transmission and detection of radiofrequency (RF) waves; provision of insights needed to solve the problem of magnetic field inhomogeneity; personalized estimation of RF power specific absorption rate (SAR); better detection of cancer; improved RF ablation and hyperthermia; and discovery of new knowledge about the structure and function of living tissue. However, despite decades of research, EP tissue mapping remains an unsolved problem.
Magnetic resonance, which can provide tomographic measurements that reflect the curvature of the electromagnetic field inside an object, is a promising tool for robust EP estimation. Our goal is to use previously untapped properties of RF coils already integral in MRI to measure electrical properties of tissue at high spatial resolution.
Our strategy comprises three different approaches. First, we solve a global inverse problem by using the integral form of Maxwell’s equations. To this end, we have introduced global Maxwell tomography (GMT), a technique for three-dimensional noninvasive EP estimation from MR measurements. Second, we implement a physics-informed neural network that solves the differential form of Maxwell’s equations. Our method, called physics-informed Fourier networks for electrical properties tomography (PIFON-EPT), estimates EP and magnetic transmit field distributions from noisy or incomplete MR measurements. Third, we have developed a supervised neural network based on a 3D vision transformer to reconstruct EP from MR measurements obtained with a birdcage coil. We are pursuing further development of these techniques with the aim of creating spatially-resolved maps of the electrical conductivity and permittivity of tissue in vivo.
Figure 1. Simulated EP reconstructions using GMT for a realistic numerical head model. We modeled an 8-channel transmit-receive 7T head coil and minimized the difference between synthetically generated noisy measurements of the coils’ transmit fields and the transmit fields estimated using updated guesses of EP during the iterative GMT inverse problem solution. The reconstructed EP closely matched the ground truth EP.
Figure 2. In vivo EP reconstructions using a neural network based on a 3D vision transformer. Electric conductivity and relative permittivity maps of a human head were estimated from the MR measurements obtained with a birdcage coil at 3T. The input included an edge mask obtained from a 3D MPRAGE image to help preserving boundaries between relevant anatomical structures.
Project Lead
This research project has been partially supported by the National Science Foundation (NSF) under Award NSF CAREER 1453675, and National Institutes of Health (NIH) under Award Number P41 EB017183.
Researchers at the Center for Biomedical Imaging at NYU Langone Health develop transformative imaging technologies to advance basic science and address unsolved clinical problems.
660 First Avenue
4th Floor
New York, NY 10016
© Center for Biomedical Imaging (CBI). All rights reserved.
This content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, NYU Langone Health, or New York University.
