Airy beam light sheet microscopy boosted by deep learning deconvolution.

Area of Science:

• Microscopy and Imaging Technologies
• Computational Biology and Neuroscience

Background:

• Standard light sheet microscopy faces limitations in balancing optical sectioning and field of view due to Gaussian beam divergence.
• Airy beams offer low divergence for improved field of view but introduce side lobes that degrade image contrast.

Purpose of the Study:

• To develop and evaluate an Airy beam light sheet microscope integrated with deep learning-based deconvolution.
• To overcome the contrast limitations caused by Airy beam side lobes and enhance imaging quality.
• To accelerate the imaging of large biological volumes, such as neural tissue.

Main Methods:

• Construction of an Airy beam light sheet microscope.
• Development of a deep learning image deconvolution method using a generative adversarial network (GAN).
• Training the GAN with high-quality data to remove side lobe effects without prior point spread function knowledge.
• Evaluation using fluorescently labeled mouse brain tissue samples.

Main Results:

• The deep learning deconvolution effectively removed side lobe artifacts from Airy beam light sheet microscopy.
• Significant enhancement in image contrast and improved performance in bicubic upscaling were achieved.
• The deep learning approach demonstrated a 20-fold speed increase compared to standard deconvolution methods.
• High-quality imaging of large volumes of fluorescently labeled neurons was successfully performed.

Conclusions:

• The combination of Airy beam light sheet microscopy and deep learning deconvolution provides a powerful solution for high-quality, large-volume imaging.
• This integrated approach overcomes the inherent trade-offs of traditional light sheet microscopy.
• The method enables rapid and high-fidelity visualization of complex biological structures like neural networks.