[Back]

A. Reismann, L. Atanasova, L. Schrangl, S. Zeilinger, G. Schütz:
"Temporal Filtering to Improve Single Molecule Identification in High Background Samples";
Molecules, 23 (2018), 333801 - 333810.

Single molecule localization microscopy is currently revolutionizing the life sciences as
it offers, for the first time, insights into the organization of biological samples below the classical
diffraction limit of light microscopy. While there have been numerous examples of new biological
findings reported in the last decade, the technique could not reach its full potential due to a set of
limitations immanent to the samples themselves. Particularly, high background signals impede the
proper performance of most single-molecule identification and localization algorithms. One option is
to exploit the characteristic blinking of single molecule signals, which differs substantially from the
residual brightness fluctuations of the fluorescence background. To pronounce single molecule signals,
we used a temporal high-pass filtering in Fourier space on a pixel-by-pixel basis. We evaluated the
performance of temporal filtering by assessing statistical parameters such as true positive rate and
false discovery rate. For this, ground truth signals were generated by simulations and overlaid
onto experimentally derived movies of samples with high background signals. Compared to the
nonfiltered case, we found an improvement of the sensitivity by up to a factor 3.5 while no significant
change in the localization accuracy was observable.

single molecule microscopy; super-resolution microscopy; Fourier filter; background fluorescence; image processing