Wide Field Deconvolution Microscopy - Note 12.3
Wide-field deconvolution microscopy consists of a tripartite system: a conventional fluorescence microscope equipped with a movable z-axis stage that permits imaging of the specimen at different focus positions, a CCD camera for quantification of the light emitted by the specimen, and a software package that is capable of correcting for distortions and information loss inherent in the imaging process. A dominant characteristic of images collected with a wide-field fluorescence microscope is that at some level the image becomes blurred. Blurring can come from two sources: 1) contributions of out-of-focus light to the imaging plane and 2) diffraction. Large scale blurring is the result of light that is reflected or emitted from objects above and below the focal plane. These contributions blur the in-focus object; generally the farther away the contribution is from the in-focus plane, the more blurred the image becomes. Diffraction is a result of the interaction of light with matter — the tendency of light waves to "bend" when they pass an obstacle. Even if an imaging system is perfectly focused, diffraction effects make it hard to discern details finer than roughly half the wavelength of the incident light. In the visible spectrum, the diffraction limit restricts resolution of objects smaller than ~200 nm (for blue light) to ~350 nm (for red light). Because many biologically interesting structures are smaller than 200 nm, the diffraction limit poses a serious limitation for optical microscopy. |
 |
| Figure 1. Bovine pulmonary artery endothelial cells were incubated with MitoTracker Red 580 (M22425) to label the mitochondria. After fixation, the cells were counterstained with DAPI (D1306, D3571, D21490) to label the nucleus. The panels show the unprocessed image (top panel), and after deconvolution (bottom panel). The image was deconvolved using Huygens software (Scientific Volume Imaging). Deconvolution was performed using Imaris software (Bitplane AG). |
Image restoration, or deconvolution, strives to correct these problems. When applied to wide-field data, deconvolution can significantly reduce blur contributions, resulting in increased resolution and greatly improved contrast 
(Figure 1./b>, Figure 2./b>). Other types of microscopy may also benefit from deconvolution: confocal laser-scanning microscopy, two-photon microscopy (
) and 4Pi microscopy. The combination of 4Pi microscopy and deconvolution is currently the pinnacle of optical resolution, allowing for resolution of objects at or below the diffraction limit.
 |
| Figure 2. FluoCells prepared slide #1 (F14780), consisting of bovine pulmonary artery endothelial cells incubated with MitoTracker Red CMXRos (M7512) to label the mitochondria. After fixation and permeabilization, the cells were stained with BODIPY FL phallacidin (B607) to label the filamentous actin (F-actin) and counterstained with DAPI (D1306, D3571, D21490) to label the nucleus. The panels show the unprocessed image (left panel), and after deconvolution (right panel). The image was deconvolved using Huygens software (Scientific Volume Imaging). Three-dimensional reconstruction was performed using Imaris software (Bitplane AG). |
The mathematical algorithms for deconvolution take a variety of forms. The earliest algorithms, developed in the mid-1980s employed techniques where the blur in each plane was suppressed by subtracting blurred versions of its "nearest neighbors." However, modern computing power has allowed far more sophisticated algorithms to become practical. For example, the fast maximum likelihood estimate (MLE) algorithm considers the light contributions made to that location by all other points within the measurement space. Doing so results in reassignment of the out-of-focus light back to the original source from which it has been estimated to have come. To improve accuracy, such algorithms function iteratively, refining the estimate of the true object in a step-by-step fashion.
Although requiring significant computing power, image deconvolution in conjunction with standard fluorescence imaging can provide significant improvements to the scientific value and quality of the images collected. Deconvolved images on this web site and in other publications make use of Huygens software (Scientific Volume Imaging), which utilizes an accelerated MLE algorithm. This software may be used for processing of time-resolved two- and three-dimensional multichannel images from wide-field, confocal, scanning disk confocal, two-photon and 4Pi microscopes. In all publications, we indicate in the picture captions when we have used deconvolution techniques to improve the resolution of the picture. Additional examples of pictures in which we have used image deconvolution include: , , , .
 |