MR-based electrical property tomography using a physics-informed network at 3 and 7 T.

Area of Science:

• Medical Imaging
• Biophysics
• Computational Biology

Background:

• Magnetic Resonance Electrical Property Tomography (MREPT) offers non-invasive in vivo electrical property (EP) quantification for tissue characterization.
• Clinical MREPT faces challenges including B1 measurement accuracy, model inaccuracies causing artifacts, and demanding hardware/software needs.

Purpose of the Study:

• To develop a novel, accurate, and high-resolution method for EP reconstruction based on water content maps using a physics-informed network (PIN-wEPT).
• To address limitations of existing MREPT techniques by utilizing standard clinical MRI protocols and hardware.

Main Methods:

• The PIN-wEPT method employs a physics-informed neural network to generate accurate water content maps, mitigating B1 field influences.
• It incorporates electrodynamic constraints from the Helmholtz equation and uses regression analysis to link water content with EPs.
• Simulations at 7T and in vivo validations at 3T and 7T on healthy subjects were performed.

Main Results:

• Numerical simulations demonstrated high accuracy with normalized mean square errors below 1.0% for water content, 11.7% for conductivity, and 1.1% for permittivity in normal brain tissues.
• In vivo studies showed good consistency of reconstructed EPs with empirical values in white matter, gray matter, and cerebrospinal fluid.
• The method effectively eliminated the influence of B1 field inaccuracies.

Conclusions:

• The PIN-wEPT method provides accurate and high-resolution electrical property reconstruction using standard clinical MRI systems.
• Its efficacy, flexibility, and compatibility suggest significant potential for future clinical applications in tissue characterization and pathology diagnosis.