High-throughput widefield fluorescence imaging of 3D samples using deep learning for 2D projection image restoration

Area of Science:

• Biomedical Imaging
• Microscopy Techniques
• Computational Biology

Background:

• Widefield fluorescence microscopy is crucial for biological process visualization.
• Conventional 3D imaging (z-stacks) is slow, causes photodamage, and generates large datasets.
• Axial z-sweep acquisition offers faster 3D imaging but produces low-quality images requiring heavy computation.

Purpose of the Study:

• To develop a novel workflow for high-throughput, high-quality 3D fluorescence imaging.
• To combine axial z-sweep acquisition with deep learning for image restoration.
• To enable quantitative analysis of complex 3D biological samples using 2D projection images.

Main Methods:

• Proposed a workflow integrating axial z-sweep acquisition with deep learning-based image restoration.
• Applied the workflow to live-cell imaging of 3D tumor spheroid cultures.
• Utilized deep learning algorithms to restore image quality from axial z-sweep data.

Main Results:

• Achieved high-fidelity 2D projection images from 3D samples.
• Demonstrated the workflow's suitability for quantitative analysis.
• Successfully imaged live-cell 3D tumor spheroids with high quality.

Conclusions:

• The combined axial z-sweep and deep learning approach enables efficient, high-quality 3D fluorescence imaging.
• This workflow overcomes limitations of traditional z-stacks for high-throughput applications.
• The method facilitates advanced quantitative analysis of complex biological structures.