Electrical Impedance Tomography Guided by Digital Twins and Deep Learning for Lung Monitoring

In recent years, there has been an increasing interest in applying electrical impedance tomography (EIT) in lung monitoring due to its advantages of being noninvasive, nonionizing, real time, and functional imaging with no harmful side effects. However, the EIT images reconstructed by traditional algorithms suffer from low spatial resolution. This article proposes a novel EIT-based lung monitoring scheme by using a 3-D digital twin lung model and a deep learning-based image reconstruction algorithm. Unlike the static numerical or experimental simulations used in other data-driven EIT imaging methods, our digital twin lung model incorporates the biomechanical and electrical properties of the lung to generate a more realistic and dynamic dataset. Additionally, the image reconstruction network (IR-Net) is used to learn the prior information in the dataset and accurately reconstruct the conductivity variation within the lungs during respiration. The results indicate that EIT using a guided digital twin and deep learning-based image reconstruction has better accuracy and anti-noise performance compared to traditional EIT. The proposed EIT imaging framework provides a new idea for efficiently creating labeled EIT data and has potential to be used in various data-driven methods for electrical biomedical imaging.