Multiscale resolution in structured illumination microscopy

Abstract

Structured illumination microscopy (SIM) is a most popular super-resolution technique used in cell biology and bio-imaging. Here, we present a novel approach to realize multiscale super-resolution SIM by swapping the non-linearity between instrumentation and reconstruction algorithm to achieve super-resolution. Our goal is to overcome two conventional limitations of SIM i.e., fixed resolution and the need of precise knowledge of illumination pattern. The optical system encodes higher order frequencies of the sample by projecting PSF-modulated binary patterns for illuminating the sample plane, which do not have clean Fourier peaks conventionally used in SIM. These patterns fold high frequency content of sample into the measurements in an obfuscated manner, which are de-obfuscated using multiple signal classification algorithm. Our approach eliminates the need of clean peaks in the illumination pattern, which have multiple advantages i.e., simple instrumentation and the flexibility of using different collection lenses. The reconstruction algorithm used in the proposed work does not require known illumination. Finally, we reduce the sensitivity of reconstruction algorithm to the signal to background ratio. Here, we acquired patterned illumination images of the same sample using different collection objective lenses, and obtained diffraction limited as well as super-resolved images, supporting 4 different resolution in the same system through SIM. Our experimental results with multiple collection objective lens show wider applicability of the proposed system at signal to background ration as small as <3.