Unveiling precision: a data-driven approach to enhance photoacoustic imaging with sparse data

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Imaging

Background:

• Photoacoustic imaging (PAI) faces challenges due to limited sensor availability and biological tissue heterogeneity.
• These limitations can result in reduced image fidelity and increased artifacts, hindering clinical applications.
• Existing methods struggle to effectively address these acquisition constraints.

Purpose of the Study:

• To introduce an innovative generative adversarial network, the Fourier Decay Perception Generative Adversarial Network (FDP-GAN).
• To address the limitations of restricted sensor availability and biological tissue heterogeneity in photoacoustic imaging.
• To enhance image fidelity and reduce artifacts in low-sampling scenarios.

Main Methods:

• Development of the Fourier Decay Perception Generative Adversarial Network (FDP-GAN).
• Integration of diverse photoacoustic data within the FDP-GAN framework.
• Utilizing generative adversarial network principles for image reconstruction and artifact reduction.

Main Results:

• FDP-GAN demonstrated significant enhancement in image fidelity.
• The network effectively reduced artifacts, particularly in low-sampling conditions.
• Improved performance was observed across diverse photoacoustic datasets.

Conclusions:

• FDP-GAN presents a novel and effective solution for improving photoacoustic imaging quality.
• The approach shows significant potential for clinical translation by overcoming acquisition limitations.
• This work marks a substantial advancement in addressing challenges in photoacoustic data acquisition.