|
|
 |
 |
Detecting Viral Infections: The above three micrographs show the difficulty of detecting infections in cell populations by optical microscopic inspection. The figure on the left illustrates uninfected baby hamster kidney (BHK) cells. The center photo was taken during early infection of the BHK cells with vesicular stomatitis virus (VSV), a pathogen of herd animals. The top right photo was taken during very late infection. Karen Duca of the Virginia Bioinformatics Institute and ECE’s Amy Bell are developing techniques to use such images to discover how host cells protect themselves from viral invaders. The images were taken at 10X magnification.
|
ECE Professor Amy Bell is applying signal and image processing techniques to microscope images in an effort to understand how viral and host cells interact. The research goal is to develop a procedure that may be used to rapidly identify viruses and the illness and mortality risks that they present.
Bell, an expert in signal analysis and reconstruction, and image compression, segmentation and analysis, is a Faculty Fellow of the Virginia Bioinformatics Institute (VBI). She is collaborating with VBI’s Karen Duca, a biophysicist studying host-virus interactions.
Detecting Cell Defenses
Duca and Bell are studying the response to infection when viruses are introduced to cells in a laboratory dish, or well. Duca is staining and measuring various protein markers that appear in response to the infection. “We are particularly interested in examining the behavior of markers in uninfected host cells for potential defensive strategies,” she said.
Defensive strategies of uninfected cells may include direct attacks on the virus, recruiting immune cells to the infection site, initiating a suicide program to prevent further viral spread, and signaling to neighboring cells that a virus is coming.
|
|
 |
Above are two montages created from low magnification epifluorescent micrographs of the MHV infection. The raw image on the left shows uneven illumination from the uv light passing through a low-power lens, as well as other signal noise. The same image after denoising is shown in the center.
|
 |
 |
|
Viral propagation signals of an influenza A virus – The top graph is from raw, uncorrected images, while the bottom is derived from images denoised by Bell’s technique. At four hours post infection, the denoised graph reveals that the infection has begun. This is not apparent in the raw images.
|
Conventional laboratory studies of viruses generally involve infecting the entire well at once. Duca is infecting the well in the center and studying the response as it spreads outward from ground zero, preserving important spatial information about how the population shares information to protect itself. She is identifying and staining relevant markers from the virus and host with chemical stains that fluoresce when viewed under uv light. Using a microscope with a low-power lens, the team then captures images of the well at regular time intervals as the infection progresses. By studying the image, they gain valuable information about innate immune responses to viruses.
Separating the Signal from the Noise
Bell’s goal is to remove the noise from these low resolution images and derive a clean immunofluorescent intensity signal (IIS). Several sources introduce noise in the images: the microscope and the fluorescent markers are the two primary sources. The microscope cannot capture the entire well at once, so at each time interval, multiple subimages must be taken quickly, then assembled in matrix fashion. Also, at low magnification, the microscope illumination is brighter in the center and dissipates toward the outer edges – creating a montage artifact in the image. “We have developed a method to remove the grid created by assembling the montage of subimages,” Bell said. “Our method – based on a model we developed that reflects the physics of fluorescence microscopy – also estimates and corrects the effect of the microscope’s uneven illumination and the markers’ spectral overlap.”
Once a montage (composite) image is denoised, it is used to derive a quantitative description of the viral propagation and host-virus interaction: this is the IIS. “The immunofluorescent intensity signals depict how the virus and host are interacting, over time, from the point of origin of the infection,” she said.
Characteristic Signals
Ultimately, the team hopes to develop a quantitative method that derives a characteristic profile – or ‘fingerprint’ – from the IIS of any host-virus system. A concurrent design goal is to develop fast methods so that laboratory results can be achieved in hours instead of days. The profiles and signal processing techniques could then be used in clinical or field settings to quickly identify known viruses, or to map unknown viruses to existing profiles to better predict their behavior and start appropriate treatment.
|
