Contextual High-Throughput 3D Volume Electron Microscopy Data Acquisition Using Artificial Intelligence

Area of Science:

• Electron Microscopy
• Artificial Intelligence
• Bioimaging

Background:

• Generating large 3D volume electron microscopy (vEM) datasets is time-consuming.
• High-resolution imaging of extensive biological structures presents significant challenges.

Purpose of the Study:

• To develop an AI-driven method to accelerate 3D vEM data acquisition.
• To introduce a dynamic-resolution scanning strategy for efficient vEM imaging.

Main Methods:

• Implemented Artificial Intelligence (AI) algorithms to generate region-of-interest masks.
• Developed an Adaptive Scanning technique for dynamic-resolution data capture.
• Applied the method across diverse biological samples, including brains, parasites, cells, and plants.

Main Results:

• Adaptive Scanning significantly reduced data acquisition times.
• The method achieved high-resolution imaging of targeted structures while capturing surrounding areas at lower resolution.
• Demonstrated sample agnosticism, proving compatibility with various organisms and tissues.

Conclusions:

• The multiresolution scanning strategy enhances throughput for 3D vEM.
• This approach offers potential time savings of up to twofold or more.
• Enables more efficient, cost-effective generation of larger vEM datasets for increased statistical power.