Deep Learning Enhanced Electrochemiluminescence Microscopy.

Area of Science:

• Biophotonics
• Computational Imaging
• Cellular Imaging

Background:

• High-quality electrochemiluminescence (ECL) imaging typically requires long exposure times (seconds), limiting high-throughput and dynamic studies.
• Image enhancement techniques are needed to improve the efficiency of ECL microscopy.
• Current methods struggle to balance image quality with rapid acquisition.

Purpose of the Study:

• To develop a deep learning strategy for reconstructing high-quality ECL images from millisecond exposure times.
• To enhance the imaging efficiency of electrochemiluminescence microscopy.
• To enable fast and information-rich imaging for dynamic chemical and biological processes.

Main Methods:

• Proposed Deep Enhanced ECL microscopy (DEECL), a general strategy utilizing artificial neural networks.
• Trained neural networks to reconstruct ECL images from short (millisecond) exposure data.
• Applied DEECL to electrochemiluminescence imaging of fixed cells and cell classification.

Main Results:

• DEECL reconstructs millisecond exposure ECL images with quality comparable to second-long exposures.
• Achieved 1 to 2 orders of magnitude improvement in imaging efficiency.
• Demonstrated 85% accuracy in cell classification using ECL data with 50 ms exposure time.

Conclusions:

• DEECL significantly enhances electrochemiluminescence imaging efficiency, enabling rapid acquisition.
• The approach facilitates data-intensive applications like cell classification with high accuracy.
• Computationally enhanced ECL microscopy is a promising tool for dynamic biological and chemical process analysis.