Sparse Coding-Enabled Low-Fluence Multi-Parametric Photoacoustic Microscopy.

Area of Science:

• Biomedical Optics
• Medical Imaging
• Microscopy

Background:

• Multi-parametric photoacoustic microscopy (PAM) enables simultaneous in vivo imaging of hemoglobin concentration, blood oxygenation, and flow speed at the microvascular level.
• High laser pulse repetition rates required for PAM limit applicable pulse energy due to safety and source constraints, impacting quantitative accuracy and visualization.

Purpose of the Study:

• To develop a novel denoising technique for PAM to enable high quantitative accuracy and structural visualization at reduced laser fluence.
• To demonstrate the effectiveness of this technique in intravital brain imaging.

Main Methods:

• A sparse coding-based two-step denoising technique was developed as an unsupervised learning approach.
• The technique was applied to multi-parametric PAM, specifically in intravital brain imaging settings.

Main Results:

• The sparse coding approach reduced laser fluence in PAM by 5 times.
• Image quality was maintained, with a structural similarity index measure (SSIM) greater than 0.92.
• Quantitative accuracy was preserved, with errors less than 4.9%.

Conclusions:

• The developed sparse coding-based denoising technique significantly relaxes laser fluence requirements for PAM.
• This approach maintains both structural imaging quality and quantitative measurement accuracy.
• It is expected to facilitate the clinical translation of multi-parametric and high-speed PAM by addressing photon budget limitations.