Learned Regularization for Microwave Tomography

2025-08-11

AI Summary

Microwave tomography (MWT) suffers from severe nonlinearity and ill-posedness, hindering conventional physics-based optimization methods from reconstructing fine-grained dielectric structures. To address this, we propose Single-Step Diffusion Regularization (SSD-Reg), which embeds a pretrained diffusion model as an unsupervised prior into a variational optimization framework—enabling physically consistent and structurally plausible reconstructions without paired ground-truth data. SSD-Reg jointly enforces data consistency under rigorous electromagnetic forward modeling and integrates a Plug-and-Play iterative scheme to incorporate strong data-driven regularization. Experiments demonstrate that SSD-Reg significantly improves reconstruction accuracy and robustness under challenging conditions—including sparse sampling and low signal-to-noise ratios—effectively mitigating the inverse problem’s ill-posedness. By unifying physics-informed modeling with learned priors in a single-step diffusion-based regularization, SSD-Reg establishes a novel paradigm for clinically viable MWT imaging.

Technology Category

• Search and Optimization: Non-convex Optimization
• Machine Learning: Optimization
• Computer Vision: Diffusion Models for Vision

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Content-based information diffusion

Abstract

Microwave Tomography (MWT) aims to reconstruct the dielectric properties of tissues from measured scattered electromagnetic fields. This inverse problem is highly nonlinear and ill-posed, posing significant challenges for conventional optimization-based methods, which, despite being grounded in physical models, often fail to recover fine structural details. Recent deep learning strategies, including end-to-end and post-processing networks, have improved reconstruction quality but typically require large paired training datasets and may struggle to generalize. To overcome these limitations, we propose a physics-informed hybrid framework that integrates diffusion models as learned regularization within a data-consistency-driven variational scheme. Specifically, we introduce Single-Step Diffusion Regularization (SSD-Reg), a novel approach that embeds diffusion priors into the iterative reconstruction process, enabling the recovery of complex anatomical structures without the need for paired data. SSD-Reg maintains fidelity to both the governing physics and learned structural distributions, improving accuracy, stability, and robustness. Extensive experiments demonstrate that SSD-Reg, implemented as a Plug-and-Play (PnP) module, provides a flexible and effective solution for tackling the ill-posedness inherent in functional image reconstruction.

Problem

• Reconstruct tissue dielectric properties from scattered fields
• Address nonlinear ill-posed inverse problem in MWT
• Improve accuracy without requiring paired training data

Innovation

• Physics-informed hybrid framework with diffusion models
• Single-Step Diffusion Regularization (SSD-Reg) for reconstruction
• Plug-and-Play module for flexible image reconstruction