Low-light Detection In Super-resolution Localization Microscopy

Super-resolution localization microscopy (or called localization microscopy) has been recognized as one of the most promising tools for life science by offering nanometers spatial resolution. Localization microscopy is essentially based on single-molecule imaging, where signal is usually as weak as a few hundred to several thousand photons and spreads onto more than ten pixels. Therefore, detection of the weak signal from single molecule is key in realizing localization microscopy. Electron Multiplying Charge Coupled Device (EMCCD) cameras are widely accepted as the image sensor for localization microscopy. However, localization microscopy is suffered from the low readout speed of EMCCD cameras (～10MHz) and provides only low imaging throughput, thus obstructing its applications in large sample imaging, such as visualizing the connectivity of neural circuits. Therefore, it is beneficial to investigate new low-light detection method for improving the imaging throughput in localization microscopy. The current thesis aims to develop quantitative methodology for discovering new low-light detection methods for localization microscopy. More details are as follows:(1) Based on the characterization of camera noise and photon transfer curves, the imaging performance of several low-light detectors was quantitatively evaluated with two key parameters including low-light detection sensitivity and image signal-to-noise ratio (SNR). The results showed that:(ⅰ) For low-light detection sensitivity, EMCCD cameras with EM gain (Andor iXon897as an example) provides highest low-light detection sensitivity (the minimum detection limit is2photon), while sCMOS cameras (Hamamatsu Flash4.0as a representative example) performs slightly lower sensitivity (the minimum detection limit is3photon),(ⅱ) For image SNR, the EMCCD camera exhibits highest image SNR under extremely weak incident signal (<13photon/pixel), while the sCMOS camera becomes the best choice with increasing signal intensity,(ⅲ) the sCMOS camera (up to400MHz for Hamamatsu Flash4.0) offers about40times higher readout rate than the EMCCD camera (～10MHz for Andor iXon897); however, sCMOS cameras suffer from poor imaging uniformity. Nevertheless, sCMOS camera seems to be a new kind of promising detector in localization microscopy by providing higher imaging speed and throughput.(2) Through repeated imaging of point-like emitters, we developed an experimental methodology to evaluate the performance of low-light detectors in single molecule detection and localization. We verified that this methodology provides～1nm measurement precision for localization precision. Utilizing this method, we demonstrated that, in the typically signal intensity range of localization microscopy (50-2000photon/pixel), a representative sCMOS camera--Hamamatsu Flash4.0provides better performance than a representative EMCCD camera--Andor iXon897. This finding provides a new detection strategy for visualizing the connectivity of neural circuits, where the imaging throughput has an increasement of about～40times.(3) Based on the single-molecule imaging model and fixed pattern noise (FPN) model, we systematically quantify the effects of FPN on single molecule localization via localization bias and localization precision was implemented. We found that:(ⅰ) FPN leads to almost no effect on localization precision, but introduces a certain amount of localization bias;(ⅱ) when additional column FPN (whose magnitude was half of the pixel FPN) was added, the localization bias increased～20%. Taking into account the true FPN magnitude in the Hamamatsu Flash4.0sCMOS (where the pixel FPN is3-5time higher than the column FPN), we verified that the effects of column FPN on single molecule localization are hard to observe. This study addresses the FPN concern which worries researchers, and thus will promote the application of sCMOS cameras in localization microscopy.