MITNet: GAN Enhanced Magnetic Induction Tomography Based on Complex CNN
Zuohui Chen, Cheng Chen, Chongyang Shao, Chang Cai, Xujie Song, Cheng Chen, Yun Xiang, Ruigang Liu, Qi Xuan
Abstract
Magnetic induction tomography (MIT) is an efficient solution for long-term brain disease monitoring. It focuses on reconstructing the brain’s bio-impedance distribution through non-intrusive electromagnetic fields. However, high-quality reconstruction of brain images remains a significant challenge, as reconstructing images from weak and noisy signals is a highly non-linear and ill-conditioned problem. In this work, we propose a generative adversarial network (GAN) enhanced MIT technique, named MITNet, based on a complex convolutional neural network (CNN). MITNet takes complex-valued signals as input and outputs a discretized conductivity distribution map. Our approach leverages the power of GANs to eliminate artifacts and enhance the reconstruction of object shapes. The experimental results on the real-world dataset validate the performance of our technique. The F1 score of MITNet surpasses the state-of-the-art SAE method by 5.33% on the agar data.