Residual-conditioned sparse transformer for photoacoustic image artifact reduction

Area of Science:

• Biomedical imaging
• Optical imaging
• Ultrasound imaging

Background:

• Photoacoustic tomography (PAT) merges ultrasound's spatial resolution with optical imaging's contrast.
• Sparse sampling in PAT reduces acquisition time and cost but introduces artifacts with conventional methods.
• Image artifacts degrade quality and diagnostic accuracy in photoacoustic imaging.

Purpose of the Study:

• To propose a novel network, Residual-Conditioned Sparse Transformer (RCST), for artifact reduction in photoacoustic images.
• To enhance image quality and diagnostic accuracy under sparse sampling conditions in PAT.
• To improve the applicability of photoacoustic imaging in biomedical research and clinical settings.

Main Methods:

• Developed a Residual-Conditioned Sparse Transformer (RCST) network.
• Incorporated residual prior information for local enhancement and detail recovery.
• Utilized sparse transformer blocks to identify and mitigate artifacts while preserving image structures.

Main Results:

• Demonstrated significant artifact suppression in simulated and experimental photoacoustic datasets.
• Showcased substantial improvements in overall image quality.
• Validated the effectiveness of the RCST network in enhancing sparse photoacoustic imaging.

Conclusions:

• The RCST network effectively reduces artifacts in sparse photoacoustic imaging.
• The proposed method enhances image quality, offering improved diagnostic potential.
• This work presents new possibilities for advanced photoacoustic imaging applications.