Our papers
Optimized molecule detection in localization microscopy with selected false positive probability
Mirek's paper about detecting molecules in microscopy images out in Nature Communications!
We present a method to detect molecules in microscopy images using probabilistic thresholding.
With only one parameter to set conditions and describe error levels, it simplifies the process
for users and enables reproducibility.
Inspired by radars and astronomy, we used signal detection theory to design an optimal filter for microscopy and
implement probabilistic thresholding that allows to control the level of false positive detections.
Why is this important? Detection of molecules is an essential first step in Single Molecule Localization Microscopy.
Yet, it currently lacks quantification of detection errors resulting in artifacts and affecting further analysis.
Furthermore, it is unduly complex.
How did we solve it? By combining two steps. First, we derived a theoretically optimal filter for Poisson
noise prevalent in microscopy, Poisson Matched Filter, and used it as a performance benchmark for other
filters used in SMLM, setting Matched Filter as a desired option. Second, we implemented probabilistic
thresholding, where the level of false positive probability sets the detection threshold and minimizes
false negative detections. To enhance robustness against varying background, we set the threshold adaptively.
We will be happy to collaborate with developers of SMLM software on implementing the method and bringing it to the users! Please,
reach out to us!
Please check our further publications at