Implicit Neural Representation for Sparse-view Photoacoustic Computed Tomography

TL;DR

This paper introduces an implicit neural representation framework for photoacoustic computed tomography that improves image quality and reduces artifacts in sparse-view imaging by modeling the initial heat distribution as a continuous function.

Contribution

The proposed INR method is the first to use a continuous neural representation for PACT reconstruction, addressing discretization errors and enhancing image fidelity in sparse-view scenarios.

Findings

INR outperforms traditional methods in artifact suppression.

INR preserves image fidelity better in sparse-view conditions.

Simulation and phantom experiments validate the effectiveness of INR.

Abstract

High-quality imaging in photoacoustic computed tomography (PACT) usually requires a high-channel count system for dense spatial sampling around the object to avoid aliasing-related artefacts. To reduce system complexity, various image reconstruction approaches, such as model-based (MB) and deep learning based methods, have been explored to mitigate the artefacts associated with sparse-view acquisition. However, the explored methods formulated the reconstruction problem in a discrete framework, making it prone to measurement errors, discretization errors, and the extend of the ill-poseness of the problem scales with the discretization resolution. In this work, an implicit neural representation (INR) framework is proposed for image reconstruction in PACT with ring transducer arrays to address these issues. pecially, the initial heat distribution is represented as a continuous function of spatial coordinates using a multi-layer perceptron (MLP). The weights of the MLP are then determined by a training process in a self-supervised manner, by minimizing the errors between the measured and model predicted PA signals. After training, PA images can be mapped by feeding the coordinates to the network. Simulation and phantom experiments showed that the INR method performed best in preserving image fidelity and in artefacts suppression for the same acquisition condition, compared to universal back-projection and MB methods.