Department of Computer Engineering
S E M I N A R
Color Graphs for Automated Cancer Diagnosis and Grading
Computer Engineering Department
In the human body, tissues are characterized with the organization of their cells. Neoplastic diseases including cancer cause changes in these organizations. A variety of graph-based computational methods have been proposed to quantify these cellular organizations for the automated cancer diagnosis and grading. These computational methods which generally use Voronoi Diagrams and Delaunay Triangulations, only consider its cell nuclei to represent a tissue, ignoring the other tissue components. Unlike these previous methods, we introduce the color graph approach that also considers different types of components for tissue representation, including luminal and stromal components. To this end, we construct a graph on all tissue components and color its edges depending on the component types of their end points. Working with the images of colon tissues, our experiments demonstrate that the features extracted from the color graphs lead to 82.65 percent test accuracy and that the color graph approach significantly improves the performance of its colorless counterparts.
DATE: 9 March, 2009, Monday@ 16:40
PLACE: EA 409