Enhancement of structural and functional photoacoustic imaging based on a reference-inputted convolutional neural...

Area of Science:

• Biomedical optics
• Medical imaging
• Computational imaging

Background:

• Photoacoustic microscopy (PAM) offers high-resolution functional imaging capabilities.
• PAM image quality is often degraded by background noise and laser biosafety limitations.
• Conventional methods to enhance PAM images can increase health risks and motion artifacts.

Purpose of the Study:

• To develop an advanced method for improving photoacoustic microscopy image quality while adhering to laser biosafety constraints.
• To introduce a novel deep learning approach, the reference-inputted convolutional neural network (Ri-Net), for noise reduction and signal enhancement in PAM.
• To validate the effectiveness and practicality of Ri-Net for microvascular imaging and functional assessments.

Main Methods:

• A reference-inputted convolutional neural network (Ri-Net) was designed and trained using photoacoustic signal and noise datasets from phantom experiments.
• The network's performance was evaluated using quantitative metrics and imaging experiments on human cuticle microvasculature.
• The trained Ri-Net was applied to multi-wavelength functional imaging of a mouse ear to assess its robustness and functional imaging capabilities.

Main Results:

• Ri-Net demonstrated significant improvement in photoacoustic signal quality.
• Quantitative analysis showed a 2.6-fold increase in image contrast and a 9.6 dB rise in signal-to-noise ratio.
• The network successfully enabled functional imaging of microvasculature, including oxygen saturation assessment in a mouse ear model.

Conclusions:

• The reference-inputted convolutional neural network (Ri-Net) effectively overcomes limitations in photoacoustic microscopy, enhancing image quality and signal-to-noise ratio.
• Ri-Net provides a practical and safe solution for high-quality microvascular imaging, improving upon conventional methods.
• This deep learning approach shows significant potential for efficient clinical microcirculation assessments and advanced functional imaging applications.