Physics-Guided Conditional Diffusion Networks for Microwave Image Reconstruction

TL;DR

This paper tackles the ill-posed electromagnetic inverse scattering problem in microwave imaging by introducing a physics-guided conditional diffusion framework that generates multiple plausible permittivity reconstructions conditioned on measured scattered fields. The method operates in a latent space learned by an autoencoder and employs a physics-aware conditioning mechanism with cross-attention to ensure data consistency, followed by a forward solver-based selection to pick the most physically plausible solution. Evaluations on synthetic and experimental data show improved robustness and generalization, with multi-frequency inputs further enhancing reconstruction fidelity and structural detail. The hybrid approach combines the generative strengths of diffusion models with explicit physical validation, offering robust, multi-solution inversions suitable for real-world microwave imaging tasks and potentially scalable to 3D medical imaging.

Abstract

A conditional latent-diffusion based framework for solving the electromagnetic inverse scattering problem associated with microwave imaging is introduced. This generative machine-learning model explicitly mirrors the non-uniqueness of the ill-posed inverse problem. Unlike existing inverse solvers utilizing deterministic machine learning techniques that produce a single reconstruction, the proposed latent-diffusion model generates multiple plausible permittivity maps conditioned on measured scattered-field data, thereby generating several potential instances in the range-space of the non-unique inverse mapping. A forward electromagnetic solver is integrated into the reconstruction pipeline as a physics-based evaluation mechanism. The space of candidate reconstructions form a distribution of possibilities consistent with the conditioning data and the member of this space yielding the lowest scattered-field data discrepancy between the predicted and measured scattered fields is reported as the final solution. Synthetic and experimental labeled datasets are used for training and evaluation of the model. An innovative labeled synthetic dataset is created that exemplifies a varied set of scattering features. Training of the model using this new dataset produces high quality permittivity reconstructions achieving improved generalization with excellent fidelity to shape recognition. The results highlight the potential of hybrid generative physics frameworks as a promising direction for robust, data-driven microwave imaging.

Figures (8)

• Figure 1: Overview of the proposed framework: (a) training phase of the denoising diffusion model; (b) inference phase for reconstructing the output from measured fields.
• Figure 2: Evaluation of the proposed model using single frequency synthetic measurements for three representative samples from each category. The best reconstruction is selected from 100 generated candidates using the proposed physics-guided selection mechanism.
• Figure 3: Four candidate reconstructions of the same scattered-field data at 5 Ghz generated from different random noise initializations, illustrating the generative and stochastic nature of the diffusion-based model.
• Figure 4: Evaluation of the proposed model on single-frequency experimental measurements, where the model was trained exclusively on synthetic data. The best reconstruction is selected from 100 generated candidates using the proposed physics-guided selection mechanism.
• Figure 5: Evaluation of the proposed model on experimental measurements, where the model was trained exclusively on using multi-frequency synthetic data generated using data from five distinct frequencies—3.0, 3.5, 4.0, 4.5, and 5.0 GHz. The best reconstruction is selected from 100 generated candidates using the proposed physics-guided selection mechanism.