by Fabian Hinterer, JKU Linz, Austria

My name is Fabian Hinterer and I will soon start my PhD in Linz as part of the SFB project “Tomography across the scales”. I was fortunate to already be involved in this project as a Masters student allowing me a small headstart into the very exciting subproject called “Ultra-high resolution microscopy”!

In single molecule microscopy, the resolution of an image is limited by the diffraction limit of the light. An example, where this becomes a problem is when we try to image small details on a cell. In particular, let’s assume we want to find out the locations of proteins on the surface of a cell. One can attach fluorescent chemical markers to these proteins and then image the emitting light. However, in the resulting microscopy image we have no chance to determine any locations! The reason is that the microscopy cannot image a point source (such as the marker attached to the protein) with pinpoint precision: The point source gets spreaded and blurred. In our example, the light from multiple emitters is merged and, making individual emitters indistinguishable. Overcoming the diffraction limit by various tricks leads to ultra-high resolution microscopy. One possible approach is to collect a large number of microscopy images, with each image containing just one (or a few) activated fluorescent markers. The idea is that the illuminated areas then don’t overlap and we can localize the proteins more easily.

Image 1: Three example images showing the non-overlapping illuminated areas

The task of getting useful microscopy images is a molecular biology problem that is handled by colleagues from TU Vienna. The step from microscopy images to obtaining localizations is then the actual mathematical problem that will be my main research area for the forseeable future. A difficulty can immediately be spotted in Image 1: It is not always completely clear if an illuminated area corresponds to signal or is just background noise. There exist various mathematical techniques to remove this background noise. Depending on the value of a threshold in this denoising procedures, we can control, how much of the background gets removed. Image 2 provides an illustration of this effect.

Image 2: Effect of different treshold values on the amount of localized proteins. The center of each circle corresponds to a localization. The larger the treshold is selected, the less proteins are found. Which value resembles reality the most closely? Ask biologists!