Sparse scanning encoding and neural network decoding for compressed photoacoustic microscopy.

Area of Science:

• Biomedical imaging
• Optical imaging technologies
• Photoacoustic microscopy

Background:

• Photoacoustic microscopy (PAM) provides high-resolution, non-invasive, label-free imaging crucial for biomedical research.
• Slow data acquisition and high sampling rates limit PAM's scalability and widespread use.

Purpose of the Study:

• To develop an efficient photoacoustic microscopy technique (IE-PAM) that overcomes current limitations in data acquisition speed and sampling requirements.
• To achieve high-quality image reconstruction from significantly reduced measurement data.

Main Methods:

• IE-PAM integrates sparse scanning encoding with neural network decoding.
• A sparse-scanning acquisition scheme uses random binary masks.
• A custom neural decoder, AFDU-Net, reconstructs images from limited measurements.

Main Results:

• IE-PAM reconstructs high-fidelity images from as little as 1.5% of the full sampling rate, a 66-fold increase in efficiency.
• Demonstrated superior performance in fine vascular fidelity, artifact suppression, and robustness in in-vivo mouse ear vasculature imaging.
• Outperformed traditional and learning-based imaging baselines.

Conclusions:

• IE-PAM minimizes information redundancy during acquisition and enables accurate reconstruction from minimal data.
• This technique lays the groundwork for efficient, fast, and scalable photoacoustic imaging in preclinical and research settings.
• IE-PAM enhances the applicability and scalability of photoacoustic microscopy for biomedical research.