Tumor Hypoxia Study - Linear Growth Model
Tumors consist of cells that vary in physiological condition. The term "hypoxic" is applied to cells within tumors that are operating with less than optimal levels of oxygen. Because tumors that are largely hypoxic are generally resistant to radiation therapy, information about hypoxia is helpful in determining the course of treatment and prognosis. The project directed by Drs. Cameron Koch and Syd Evans is aimed at using computer-aided image analysis and related statistical techniques in rapid determination of tumor condition from images of tumor sections or of disassociated cells. Images are recorded electronically to show the results of several fluorescent staining techniques. One of these is an experimental compound called Ef5. This stain is administered in-vivo, and binds more readily to cells as the oxygen tension decreases. As a result, the more hypoxic the cell, the more brightly it should glow. When cells are disassociated and viewed in a dark-field, they appear as shown below.
The image is one that was made from cells grown in tissue culture. Our eyes do a very good job of separating the brighter cells from the dim cells and both from the variable background, and of determining counts of cells in various conditions. However, automating the procedure is another matter. On a flat background, we could screen for all objects with a particular intensity level. However, we have major trends in the background intensity, meaning that a given light intensity value from a cell does not mean the same thing at all points in the field. We also have a great deal of noise in the background and within objects. The result is that the intensity ranges for bright cells, dim cells, and background overlap. In addition, if the cells had come from a real biopsy, the image would have had much more variability in cell size, and there would have been much more debris, consisting of particles of various sizes, shapes, and intensities. The goal of our work is to develop algorithms for processing the images, separating the cells from background and debris, and obtaining the distribution of Ef5 staining among those cells.