3-D image restoration of fluorescence microscope images by regularization

Abstract

By combining optical sections of a fluorescently labelled cell with quantitative calibration of the microscope’s blurring, we are able to computationally reverse the blurring of the microscope. We apply this approach to images taken on a conventional wide field microscope using a cooled CCD camera. Our algorithm is flexible in its use of data, which enables us to obtain good results with a small number of optical sections. We use this flexibility to minimize photodamage and photobleaching, so that we are able to obtain 3-D movies of living cells with high spatial and temporal resolution. We obtain rejection of out of focus light, resolution and accuracy of quantitative fluorescent measurement superior to confocal microscopes with substantially less photobleaching and photodamage to the sample. Recent algorithmic advances have enabled us to achieve transverse resolution as good as 100nm and axial resolution as good as 400nm.The wide variety of cells and labels examined in 3-D in our lab illustrate the robustness and effectiveness of our image restoration approach. These applications range from samples as sparsely labelled as single copy genes in lymphocytes and fibroblasts labelled by in situ hybridization to densely labelled cells with fluorescently labelled dextran or Fura-2 that completely fill the cell volume.