Accelerated model-based iterative reconstruction strategy for sparse-view photoacoustic tomography aided by...

Area of Science:

• Biomedical Imaging
• Medical Physics
• Computational Imaging

Background:

• Sparse-view data acquisition in photoacoustic tomography (PAT) is often necessary due to hardware limitations.
• Traditional reconstruction algorithms struggle with sparse-view PAT, leading to significant artifacts and image deterioration.

Purpose of the Study:

• To develop an accelerated model-based iterative reconstruction strategy for sparse-view PAT.
• To enhance reconstruction quality by incorporating multi-channel autoencoder priors.

Main Methods:

• A multi-channel denoising autoencoder network was designed to learn prior information for image reconstruction.
• The learned priors were integrated as constraints into a model-based iterative reconstruction framework.
• The strategy was validated using both simulated blood vessel data and in vivo experimental data.

Main Results:

• The proposed method achieved superior sparse-view PAT reconstruction with accelerated iteration times.
• Significant reduction in artifacts and improved image quality were observed compared to traditional methods.
• Under extremely sparse conditions (32 projections), the method showed a 48% PSNR and 12% SSIM improvement over U-Net for in vivo data.

Conclusions:

• The novel accelerated reconstruction strategy effectively addresses sparse-view limitations in PAT.
• The integration of multi-channel autoencoder priors enhances reconstruction accuracy and speed.
• This AI-aided approach offers a promising solution for high-quality sparse-view photoacoustic imaging.