Blind Deconvolution of Low and High Signal-to-noise 3-D Images of Fluorescent Subcellular Structures

Abstract

This report summarizes our experiences gained with deconvolving 3-d images of various types of biological samples - animal and plant cells; live and fixed. We compared raw, unmodified images with images processed with standard filtering methods and images improved by using deconvolution. The blind decovolution software package from AutoQuant, Inc. was used. All images were acquired using a BioRad MRC 1024 confocal microscope (attached to a Nikon Diaphot microscope), using fluorescence or reflected light. For thin (up to 25 μm) flat specimens a 60x PlanApo NA1.4 lens was used while for thick (25-200 μm) water containing samples a 40x Fluor NA1.15 water immersion lens was used.In an attempt to test the usefulness and validity of deconvolution for various types of biological samples we adopted the following approaches:1. collect images of structures that are known to be both fluorescent and reflective - compare the results of deconvolution of fluorescence images with reflected light images that are not deconvolved.