Feasibility study for conductivity reconstructions from spin-echo images using artificial neural network with simulation data in 3T MR system

Abstract

Phase-based Electrical properties tomography is a non-invasive imaging technique that uses MRI systems to measure the tissue conductivity. However, the conductivity reconstruction process causes problems such as noise amplification and boundary artifact. To address such limitations, several DL-based reconstruction methods were proposed. Building upon these works, we propose an ANN-based conductivity reconstruction method trained only on simulation dataset. The proposed method was studied with the aim of: (a) approaching ground-truth conductivity values, (b) noise-robustness, (c) higher image resolution, (d) generalization to clinical data. The feasibility was investigated on simulations and TSE in-vivo data (one healthy volunteer, two meningioma cases).