A simulated example for co-localisation

Higher resolution can always be considered as an advantage. However, it can change the view on standard image analysis concepts, as for example measuring co-localisation of fluorescently labelled proteins. The figure_displays the same simulated structure (ground truth) as imaged at different resolution. In addition, it reports the commonly used Pearson coefficient as one possible measure for co-localisation: In a standard confocal image the structure seems to be partially co-localised which is also reflected by a Pearson coefficient of 0.6. By applying deconvolution or other means of increasing the resolution moderately into a range of 140 nm, the Pearson coefficient drops to 0.17 and the structure seems to consist of two concentric rings with different diameters. If nanoscopy (e.g. STED microscopy) is applied with a resolution of 30 nm, the rings can be subdivided into dots that do not co-localise at all (Pearson coefficient: 0.0).

The concept of co-localisation should therefore be always seen (and reported) as being dependent on the resolution applied in the imaging process. It needs to be noted that the two fluorophore-labelled proteins can of course never occupy the exact same space and that they only appear to do so due to the resolution limit and subsequent blurring coming from the imaging technology applied.

Microscopy image simulation of red and green diffraction-limited spots organized in two rings as seen at different levels of resolution, and the effect of resolution on co-localisation analysis (Object Pearson coefficient). Image simulation and deconvolution powered by Huygens Professional. Image used with permission. Courtesy of Remko Dijkstra, Scientific Volume Imaging (www.svi.nl).