Kernel-based regularization for photoacoustic pressure reconstruction

Abstract

Photoacoustic tomography (PAT) is a promising imaging technique that combines the high spatial resolution of ultrasound with the high contrast of optical imaging. One of the challenges in PAT is the ill-posed nature of the inverse problem, where limited measurement data and noise often lead to inaccurate reconstructions. This work introduces a kernel-based regularization (KBR) approach for model-based reconstruction algorithms in photoacoustic (PA) imaging. The proposed method leverages kernel-induced feature space to enforce smoothness and spatial coherence in the reconstructed images, thereby improving the robustness to noise and data sparsity. By incorporating prior knowledge of the signal dynamics for solving the PA inverse problem, KBR enhances the reconstruction fidelity, especially in regions with low signal-to-noise ratio. Numerical experiments and phantom studies demonstrate that the proposed algorithm outperforms traditional regularization techniques, such as Tikhonov and total variation regularization, regarding reconstruction accuracy and computation speed. The results suggest KBR provides a powerful tool for addressing the inherent challenges in PA image reconstruction, offering potential improvements in several applications.