Depth-extended acoustic-resolution photoacoustic microscopy based on a two-stage deep learning network

Area of Science:

• Biomedical optics
• Medical imaging
• Photoacoustic imaging

Background:

• Acoustic resolution photoacoustic microscopy (AR-PAM) offers high-resolution imaging of biological tissues.
• AR-PAM image quality degrades significantly with increasing depth due to limited depth-of-focus.
• Existing methods to extend imaging depth often have stringent focal point requirements not met by standard AR-PAM systems.

Purpose of the Study:

• To develop and validate a novel deep learning (DL) strategy for adaptive, high-resolution AR-PAM image reconstruction at extended depths.
• To overcome the limitations of conventional AR-PAM in imaging targets outside the focal plane.
• To enhance the practical utility of AR-PAM for deeper biomedical investigations.

Main Methods:

• A two-stage deep learning reconstruction strategy was proposed, utilizing a residual U-Net with an attention gate.
• The DL network was optimized through phantom and in vivo experiments.
• The method reconstructs high-resolution photoacoustic images adaptively for varying out-of-focus depths.

Main Results:

• The proposed DL method successfully extended the effective depth-of-focus of AR-PAM from 1 mm to 3 mm.
• Imaging resolution in regions 2 mm beyond the focal plane was significantly improved, approaching in-focus quality.
• The system maintained high imaging performance under a light energy density of 4 mJ/cm².

Conclusions:

• The developed DL reconstruction strategy effectively enhances the imaging depth and resolution of AR-PAM.
• This advancement broadens the applicability of AR-PAM in biomedical studies requiring visualization of deeper tissue structures.
• The method offers a promising solution for improving the capabilities of current AR-PAM systems.