Optimizing quantitative photoacoustic imaging systems: the Bayesian Cramér-Rao bound approach.

Area of Science:

• Medical Imaging
• Computational Science
• Biophysics

Background:

• Quantitative photoacoustic computed tomography (qPACT) offers high-contrast, high-resolution medical imaging.
• qPACT image reconstruction involves complex, non-linear, ill-posed inverse problems governed by PDEs.
• Lack of standardized qPACT system designs hinders optimal application development.

Purpose of the Study:

• To develop a novel computational approach for optimal experimental design of qPACT systems.
• To address challenges in applying Bayesian Cramér-Rao bounds in infinite-dimensional settings for qPACT.
• To create a computationally efficient and estimator-independent design metric.

Main Methods:

• Utilized the Bayesian Cramér-Rao bound (CRB) for optimal experimental design.
• Incorporated techniques for infinite-dimensional function spaces, including trace-class covariance priors.
• Employed the variational adjoint method for computing log-likelihood derivatives.

Main Results:

• Introduced a novel Bayesian CRB-based design metric for qPACT systems.
• The proposed metric is computationally efficient and independent of the chosen inverse problem estimator.
• Demonstrated the metric's efficacy in guiding experimental design through a 2D numerical study.

Conclusions:

• This work presents the first Bayesian CRB-based design approach for PDE-governed systems like qPACT.
• The developed method provides a framework for optimizing qPACT system design.
• This facilitates the advancement of qPACT technology for improved medical imaging applications.