Neural network-assisted localization of clustered point spread functions in single-molecule localization microscopy.

Area of Science:

• Biophysics
• Microscopy Techniques
• Computational Biology

Background:

• Single-molecule localization microscopy (SMLM) enables nanoscale imaging but struggles with fluorophore clustering in dense samples, reducing localization accuracy.
• Fluorophore clustering is a significant limitation for achieving high-resolution SMLM, hindering detailed analysis of biological structures.

Purpose of the Study:

• To develop and validate a novel convolutional neural network (CNN)-assisted approach for accurate localization of clustered fluorophores in SMLM.
• To improve the precision and resolution of SMLM imaging, particularly in challenging, densely labeled biological samples.

Main Methods:

• A convolutional neural network (CNN) was trained on simulated SMLM images to predict point spread function (PSF) locations by generating Gaussian blobs.
• Blob detection was utilized as a post-processing step to refine predicted PSF locations and enhance localization precision.

Main Results:

• The proposed CNN-assisted approach significantly improved PSF localization accuracy compared to traditional methods, especially in densely labeled samples.
• Post-processing with blob detection further refined localization precision, demonstrating the combined efficacy of the CNN and blob detection techniques.

Conclusions:

• Convolutional neural networks are highly effective in overcoming fluorophore clustering challenges in SMLM.
• This approach advances SMLM spatial resolution, offering deeper insights into the intricate details of complex biological structures.