Deep-learning-based whole-brain imaging at single-neuron resolution.

Area of Science:

• Neuroscience
• Biomedical Imaging
• Computational Biology

Background:

• Understanding neural circuits requires high-resolution 3D imaging of neuronal structures.
• Current methods for large-scale, optical-resolution whole-brain imaging are often complex and inefficient.
• Advanced imaging techniques are needed to overcome the limitations of existing technologies.

Purpose of the Study:

• To develop a novel deep-learning-based method for high-throughput, high-resolution whole-brain imaging.
• To enable efficient visualization of fine neuronal structures across the entire brain.
• To provide a robust tool for neuroscience research and brain function studies.

Main Methods:

• Developed a deep-learning-based fluorescence micro-optical sectioning tomography (DL-fMOST) approach.
• Employed a U-net convolutional neural network for real-time optical sectioning.
• Integrated wide-field microscopy and histological sectioning to extend imaging depth.

Main Results:

• Successfully acquired a 3D whole-mouse brain dataset with a voxel size of 0.32 × 0.32 × 2 µm.
• Achieved high-throughput imaging, completing data acquisition in just 1.5 days.
• Demonstrated the method's robustness across mouse brains with various neuron labeling types.

Conclusions:

• DL-fMOST offers a powerful solution for high-resolution, large-scale whole-brain imaging.
• The method significantly advances the ability to study complex neuronal architectures.
• This technique is a valuable tool for investigating brain function at the structural level.