Localisation microscopy using quantum dots

Localisation microscopy (LM) techniques such as STORM/fPALM have proved to be viable techniques in biological research. These techniques provide super-resolution images of fluorescently labelled biological samples by localisation of
individual fluorophores. LM techniques become increasingly popular choice for
investigation of detailed structure of biological samples because of their
inexpensive realisation. We are working on an LM related technique using quantum dots (QD) as fluorescent markers. Quantum dots - with an order of magnitude higher brightness then organic fluorophores, exceptional photostability and non-toxicity - seem to be ideal fluorophores for fluorescence microscopy study of biological samples. Under a constant illumination, QD exhibit stochastic blinking in emission of fluorescent light. We use the fluorescence intermittency to separate individual overlapping QD within a diffraction limited volume. This makes the precise localisation of individual QD possible, and sub-resolution structure of the sample can be revealed. We record a time stack of wide field images of the sample stained with blinking QDs. To separate individual sources (QDs) we use a model based on non-negative matrix factorisation (NMF). Our NMF algorithm is a natural model for data corrupted with Poisson noise (such as microscopic images) using Czisar's I-divergence as a cost function. NMF based methods do not require knowledge about the point spread function (PSF) which makes it possible to use for 3D samples where fluorophores can be above or below the plane of focus or can be corrupted with various aberrations changing the shape of PSF across the field of view. We hope the method can significantly speed up the acquisition time for LM techniques, as well as make the use of QD possible for super-resolution fluorescence