Deep-learning-based motion-correction algorithm in optical resolution photoacoustic microscopy.

Area of Science:

• Biomedical imaging
• Deep learning
• Photoacoustics

Background:

• Motion artifacts degrade image quality in optical resolution photoacoustic microscopy (OR-PAM).
• Accurate artifact correction is crucial for analyzing OR-PAM data, especially in vivo.
• Existing methods may lack flexibility for diverse motion scenarios.

Purpose of the Study:

• To develop and validate a deep learning-based method for correcting motion artifacts in OR-PAM.
• To assess the method's performance on both simulated and in vivo data.
• To demonstrate the adaptability of the proposed method for various OR-PAM applications.

Main Methods:

• A convolutional neural network (CNN) was designed for end-to-end artifact correction.
• The CNN was trained and evaluated using simulated OR-PAM data.
• The method was applied to in vivo OR-PAM images of rat brain vasculature.

Main Results:

• The deep learning method effectively corrected motion artifacts in simulated OR-PAM data.
• In vivo results showed successful artifact correction in rat brain vessel images, including capillaries.
• The method demonstrated robustness for both large vessels and fine capillary networks.

Conclusions:

• The proposed deep learning approach offers a powerful solution for motion artifact correction in OR-PAM.
• The method is adaptable to different motion correction needs by adjusting training datasets.
• This technique enhances the utility of OR-PAM for in vivo imaging applications.