Resolution enhancement of wide-field interferometric microscopy by coupled deep autoencoders

Area of Science:

• Biomedical imaging
• Nanotechnology
• Artificial intelligence

Background:

• Wide-field interferometric microscopy (WFI) offers sensitive, label-free imaging of nanoparticles.
• Current WFI resolution limits hinder accurate detection and classification of subdiffraction-limited biological entities like viruses and exosomes.

Purpose of the Study:

• To develop a deep learning approach for enhancing the resolution of WFI images.
• To improve the accuracy of nanoparticle detection and classification by overcoming resolution limitations.

Main Methods:

• A deep learning model utilizing coupled deep autoencoders was designed.
• The network was trained using paired microscope image patches and corresponding ground truth image patches.
• The trained network learns to transform low-resolution input images into high-resolution outputs.

Main Results:

• The proposed deep learning method successfully enhanced the resolution of images of L-shaped nanostructures.
• The network demonstrated the ability to reconstruct denoised and resolution-enhanced image patches from unseen input data.
• This approach holds potential for improving nanoparticle visualization and analysis.

Conclusions:

• Deep learning, specifically coupled deep autoencoders, can effectively enhance the resolution of wide-field interferometric microscopy images.
• This method offers a promising solution for overcoming resolution challenges in nanoparticle imaging.
• The developed technique can advance the accurate detection and classification of nanoscale biological structures.