Localizing axial dense emitters based on single-helix point spread function and deep learning

Yihong Ji,Danni Chen,Hanzhe Wu,Gan Xiang,Heng Li,Bin Yu,Junle Qu

2024-03-05

Abstract:Stimulated Emission Depletion Microscopy (STED) can achieve a spatial resolution as high as several nanometers. As a point scanning imaging method, it requires 3D scanning to complete the imaging of 3D samples. The time-consuming 3D scanning can be compressed into a 2D one in the non-diffracting Bessel-Bessel STED (BB-STED) where samples are effectively excited by an optical needle. However, the image is just the 2D projection, i.e., there is no real axial resolution. Therefore, we propose a method to encode axial information to axially dense emitters by using a detection optical path with single-helix point spread function (SH-PSF), and then predicted the depths of the emitters by means of deep learning. Simulation demonstrated that, for a density 1~ 20 emitters in a depth range of 4 nm, an axial precision of ~35 nm can be achieved. Our method also works for experimental data, and an axial precision of ~63 nm can be achieved.

Optics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to effectively obtain the three - dimensional position information of axially dense emitters when using non - diffracting optical needles to excite samples. Specifically, the paper proposes a method that combines the Single - Helix Point Spread Function (SH - PSF) and deep learning (named SH - DL) to achieve high - precision positioning of axially densely distributed emitters. When dealing with axially dense emitters, traditional methods are difficult to accurately extract the position information of each emitter due to the severe overlap of spots in the image. By using SH - PSF to encode the emitter information at different depths and using deep - learning algorithms to analyze these overlapping images, this method can significantly improve the accuracy and density of axial positioning.

The main technical challenges mentioned in the paper include:
1. **Encoding of axial information**: Use the SH - PSF detection path so that emitters at different depths appear as different azimuths in the image, thus encoding the axial information.
2. **Analysis of overlapping images**: Use the encoder - decoder network architecture in deep learning to accurately predict the position of each emitter from the overlapping images.
3. **Handling of high - density emitters**: In high - density situations, traditional single - molecule localization methods are difficult to distinguish overlapping spots, while deep - learning methods can effectively handle such high - density situations.

Through simulation and experimental verification, this method shows high positioning accuracy under different densities of emitter distributions, especially in the axial direction. This provides a new solution for quickly obtaining volume images with 3D nanometer resolution.