Light & Fluorescence Microscopy
Image reconstruction from SMLM data
The result of the single molecule localisation routine e.g. via Gaussian-fitting is not an image, but a table with the coordinates of the events. An event is the fluorophore localised in a given frame. Please note that a fluorophore can emit photons in multiple frames and therefore can give raise to multiple events. The localisation precision (derived from the number of photons) is reported in the table amongst some other parameters for data handling (e.g. an event ID).
This table needs to be converted into an image or in other words the super resolution image needs to be reconstructed. Two common approaches will be briefly described here. One approach is the histogram (based) reconstruction. In this case, a new image is created with a smaller pixel-size than the original camera image to account for the increased resolution in the super-resolution image. Typically, a pixel size of 20 nm is chosen, but the operator of the software is free to use other values (to match the Nyquist criterion) on the one hand and still retain a usable S/N ratio). The algorithm is allocating every event to a pixel just by determining if the x, y, (z) position is within the area (volume) covered by the pixel (voxel). The number of events per pixel is then displayed as the brightness of the pixel.
The Gauss-reconstruction can be explained easiest in a slightly simplified manner: The algorithm turns every event into a gauss function representing the localisation precision (sigma) by its width. The 2D Gaussian-function for each event is then placed at the x, y position of the event in the reconstructed image. Finally, an (adjusted) sum projection is performed to create the final image. Usually, a pixel size of 5 nm is selected to be able to appreciate the details of the Gaussian-shape of the events reconstructed.
![image description](images/LM-module/figure_79.jpg)