Computational cannula microscopy of neurons using neural networks

Area of Science:

• Biomedical Engineering
• Microscopy
• Artificial Intelligence

Background:

• Computational cannula microscopy offers minimally invasive, high-resolution imaging deep within tissues.
• Current methods face limitations in real-time processing and scalability for larger fields of view.

Purpose of the Study:

• To develop real-time, power-efficient image reconstruction for computational cannula microscopy using artificial neural networks.
• To enhance the scalability of imaging techniques for broader applications.

Main Methods:

• Application of artificial neural networks for image reconstruction.
• Demonstration using widefield fluorescence microscopy on cultured neurons and fluorescent beads.
• Utilizing a narrow cannula (220 µm diameter) for deep tissue imaging.

Main Results:

• Achieved real-time, power-efficient image reconstructions.
• Demonstrated high-resolution imaging (<10 µm) with a wide field of view (200 µm diameter).
• Successfully extended the approach to macro-photography.

Conclusions:

• Artificial neural networks significantly improve computational cannula microscopy.
• The enhanced technique offers scalable, high-resolution imaging capabilities for biological and other applications.
• This advancement opens new possibilities for in-situ tissue analysis and imaging.