Improving Resolution in Wide-field Fluorescence Microscopy Using Deconvolution Techniques

Fluorescence microscopy has been instrumental in studying complex biological functions by indicating specific regions of interest in live cell imaging. This is usually used in conjunction with the non-invasive technique of optical sectioning to image multiple focal planes within a thick sample, allowing its 3D reconstruction during post-processing. However, since objects outside the focus plane also receive illumination, they too fluoresce and interfere with the images captured. This unintended interference drastically affects the resolution of the images along the optical axis (z-axis), resulting in severely blurred 3D reconstructions.  Though there have been several attempts to  alleviate this problem with instrument modifications, the focus in the work presented is on investigating the use of computational reconstruction techniques, commonly known as deconvolution. Assuming a linear shift-invariant imaging model, the blur introduced during optical sectioning is characterized by the distortion introduced to a point source in the specimen, referred to as the point spread function (PSF). Using fluorescent beads to approximate point sources, a comparison study between the measured PSF and theoretical models, generated according to the experimental parameters, is presented. The PSF models are then employed with multiple state-of-the-art deconvolution algorithms to deblur 3D datasets of a zebrafish embryo acquired through optical sectioning in wide-field microscopy. Finally, a multi-view deconvolution technique is demonstrated with simulated datasets, acquired from multiple-angles about an axis orthogonal to the z-axis, yielding better reconstructions than single-view deconvolution.