Bilkent University
Department of Computer Engineering

---

S E M I N A R

 

The Cell-Graphs of Cancer

 

Cigdem Gunduz Demir

Department of Computer Science
Rensselaer Polytechnic Institute

In the current practice of medicine, pathologists traditionally diagnose cancer from tissue samples. Examining such biopsy samples under a microscope, a pathologist typically makes assessments based on visual interpretation of cell morphology and tissue distribution. This, however, leads to a certain level of subjectivity, possibly resulting in some inter-observer variability. To circumvent this problem, it is important to develop computational diagnostic tools that operate on quantitative measures. Such automated diagnostic tools facilitate fast, objective, mathematical judgment complementary to that of a pathologist, reducing the subjectivity. For the purpose of automated cancer diagnosis, we introduce a new computational cancer model based on the generation of cell-graphs from tissue samples. In this approach, we employ machine learning algorithms to automatically distinguish cancerous tissues from their counterparts by making use of the distinctive topological properties of “cluster” formation in cancerous cells. In this talk, we will present the methodology of cell-graph generation along with a theoretical framework and experimental demonstrations. We will introduce the definitions of different sets of distinctive cell-graph features (such as clustering coefficient, giant connected component ratio, spectral radius, and number of connected components). In the talk, we will also report on the experimental demonstrations obtained on clinical data for the diagnosis of brain cancer (glioma). Despite the complex dynamic nature of glioma formation, we successfully demonstrate that the self-organizing clusters of cancerous cells in human brain exhibit distinctive local and global graph properties and, hence, that a machine learning algorithm (such as an artificial neural network and a support vector machine) is able to differentiate the cancerous tissue from non-cancerous tissues, for example, from a healthy tissue or a benign inflammatory process (an “inflamed” tissue) with high accuracies.

 

DATE: June 13, 2005, Monday @ 13:40
PLACE: EA 409