Optimizing Super-resolution Localization Microscopy: Three Dimensions, Dense Scenes, Fast drift-less acquisition, and Multicolor/Multispectral Imaging
University of Colorado Boulder, USA
The advent of fluorescence microscopy over the last decades has revolutionized biology. However, the use of optical radiation limits the resolution of fluorescence images to approximately one half the wavelength of the light used in imaging (a few hundred nanometers). Many sub-cellular objects are one or two orders of magnitude smaller than this diffraction limit. To overcome this limitation, one can impose the condition that only a sparse subset of the fluorescent molecules are active in any given image frame, thus enabling each molecule to be resolved and individually localized to very high precision. Localizations from thousands of frames are then combined computationally, yielding an image with resolution dictated by the precision with which fluorescent molecules are localized. By modifying the point spread function of the imaging path of the microscope using a photon-efficient phase mask, we can collect high-precision information regarding the axial location of the fluorophores as well, thus extending the localization microscopy to 3D. This methodology is extended to multicolor/multispectral imaging. We further introduce sparsity-based algorithms for identifying overlapping images of fluorescent molecules. By relaxing the stipulation that the images of the individual molecules be well separated, we can identify more molecules per frame of data. Our algorithms increase the identifiable density of emitters by a factor of ten, resulting in a comparable decrease in data collection time. When applied to biological data, we are able to compute full 3D reconstructions of objects with feature sizes more than an order of magnitude smaller than the diffraction limit.