| dc.description.abstracteng | Super-resolution fluorescence techniques enable the study of structures smaller than the diffraction
limit with visible light microscopy. Their introduction and development within the last
two decades opened up the possibility to ask entirely new questions in cell biology. The field of superresolution
microscopy is still growing rapidly, and many improvements and novel methods have been
proposed in recent years.
The major part of this thesis revolves around the advancement of Super-resolution Optical Fluctuation
Imaging (SOFI), a fairly new technique which enhances the spatial resolution of an image by evaluating
the temporal fluctuations of blinking fluorescent emitters. SOFI enables optical sectioning with
wide-field microscopes and is compatible with a large range of experimental conditions. The following
SOFI-related contributions are presented: First, a comprehensive analysis of the convergence properties
of SOFI and its dependence on experimental parameters is shown, including an estimation of
the necessary recording time and particle density for a targeted resolution enhancement. Second, we
present Fourier SOFI, a new approach to generate super-resolved images on a finer pixel grid than the
original camera recording. In contrast to established algorithms relying on spatial cross-cumulants, this
method is practically free of artifacts and does not require any post-processing corrections. Next, an
algorithm is laid out that corrects for the contributions of noise in zero-time-lag SOFI images. This
extends the applicability of auto-cumulant SOFI – which correlates values only in time, not in space
– to recordings where the time scale of photoblinking is on the same order as the exposure time. We
also show how the lateral microscope Point Spread Function (PSF) can be estimated from SOFI data
of thin samples. If multiple focal planes are imaged at the same time, the full three-dimensional PSF
can be recovered. The fifth contribution revolves around Fourier Preweighting, a method that increases
the resolution of SOFI images by pre-processing the original data. We show that this outperforms current
techniques and surprisingly also improves the density dependence of SOFI. A related algorithm
is presented which automatically matches the degree of resolution enhancement to the data quality, avoiding artifacts. The last SOFI contribution demonstrates its applicability to cells labeled with carbon
nanodots, a fairly new cost-effective and bio-compatible class of fluorophores, and shows that their
blinking behavior is qualitatively similar to that of fluorescing semiconductor crystals (quantum dots).
Unrelated to SOFI, we show how single Atto647N molecules can be localized with sub-nanometer
precision by exploiting the increased photon yield of samples at liquid nitrogen temperature. We propose
a method to resolve spectrally identical, non-blinking fluorophores spaced only few nanometers
apart with a special cryo-fluorescence setup and validate our concept with simulations.
We also developed TrackNTrace, an open-source MATLAB framework to support the development
of fluorescence imaging applications. Its design is focused on easy extensibility through plugins, simplicity
of coding, and rich visual feedback. We demonstrate competitive performance and execution
speed in Single-Molecule Localization Microscopy applications compared to established software and
include many state-of-the-art algorithms out-of-the-box.
Finally, we present two projects reliant on single-molecule imaging: First, we developed a model
for the intensity distribution of fluorescent molecules imaged while flowing through a nanochannel and
verified it experimentally. This can be used to extract the ratios of differently labeled species from a measurement
of their mixture. In the second project, we demonstrate the first simultaneous measurement
of the excitation and emission dipole axes of single molecules. | de |