Diff-INR: Generative Regularization for Electrical Impedance Tomography

Bowen Tong,Junwu Wang,Dong Liu

2024-09-10

Abstract:Electrical Impedance Tomography (EIT) is a non-invasive imaging technique that reconstructs conductivity distributions within a body from boundary measurements. However, EIT reconstruction is hindered by its ill-posed nonlinear inverse problem, which complicates accurate results. To tackle this, we propose Diff-INR, a novel method that combines generative regularization with Implicit Neural Representations (INR) through a diffusion model. Diff-INR introduces geometric priors to guide the reconstruction, effectively addressing the shortcomings of traditional regularization methods. By integrating a pre-trained diffusion regularizer with INR, our approach achieves state-of-the-art reconstruction accuracy in both simulation and experimental data. The method demonstrates robust performance across various mesh densities and hyperparameter settings, highlighting its flexibility and efficiency. This advancement represents a significant improvement in managing the ill-posed nature of EIT. Furthermore, the method's principles are applicable to other imaging modalities facing similar challenges with ill-posed inverse problems.

Image and Video Processing,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper aims to address the reconstruction problem in Electrical Impedance Tomography (EIT). EIT is a non-invasive imaging technique that applies small currents through electrodes placed on the body surface and measures the resulting voltages to reconstruct the internal conductivity distribution. However, EIT reconstruction faces a severely ill-posed nonlinear inverse problem, making accurate results difficult to obtain.

To tackle this challenge, the paper proposes the Diff-INR method, a novel approach that combines generative regularization with Implicit Neural Representations (INR). Specifically:

1. **Generative Regularization**: By introducing a diffusion model, Diff-INR can utilize geometric priors to guide the reconstruction process, effectively addressing the shortcomings of traditional regularization methods.

2. **Implicit Neural Representations**: By integrating a pre-trained diffusion regularizer with INR, this method achieves state-of-the-art reconstruction accuracy on both simulated and experimental data.

3. **Flexibility and Efficiency**: Diff-INR performs excellently under different grid densities and hyperparameter settings, highlighting its flexibility and efficiency.

Moreover, this method is not only applicable to EIT but also to other imaging modalities facing similar ill-posed inverse problems. In summary, Diff-INR significantly improves the accuracy, robustness, and generalization capability of EIT reconstruction by introducing generative regularization.