Image reconstruction with uncertainty quantification in photoacoustic tomography

Area of Science:

• Biomedical Imaging
• Medical Physics
• Computational Imaging

Background:

• Photoacoustic tomography (PAT) is a hybrid imaging modality.
• PAT combines optical contrast with ultrasound resolution.
• The core challenge is reconstructing initial pressure distributions from detected ultrasound waves.

Purpose of the Study:

• To describe a Bayesian approach for solving the inverse problem in photoacoustic tomography.
• To detail the computation of point estimates for image reconstruction.
• To enable uncertainty quantification in PAT.

Main Methods:

• Developed a Bayesian framework for photoacoustic tomography.
• Derived the solution for the inverse problem.
• Investigated the approach using simulations with varying detector geometries (limited view) and properties (point-like, finite size, finite bandwidth).

Main Results:

• The Bayesian approach provides accurate estimates of the initial pressure distribution.
• The method successfully quantifies the uncertainty associated with these estimates.
• Simulations demonstrated robustness across different detector configurations and properties.

Conclusions:

• The Bayesian approach is effective for accurate image reconstruction in photoacoustic tomography.
• This method offers valuable uncertainty quantification for PAT.
• The approach shows promise for improving diagnostic capabilities in photoacoustic imaging.