Browsing all articles in publications
Mar
17

BSEC 2011 Poster

Author Robert Coop    Category publications     Tags

A poster that I presented at the 3rd annual Oak
Ridge National Lab Biomedical Science and Engineering Conference.

Functional analysis and prediction of tumor growth

Abstract:

Based on a modified logistic model, simulated growth trajectories of brain tumors are prepared. Simulation trajectories track the number of cancer cells within each 3-dimensional area (voxel) in a 128×128 grid. A novel functional analysis procedure is developed which uses a genetic algorithm combined with systems theory principles in order to determine, for a given voxel, which neighboring voxels have the greatest predictive power when compared to the neighborhood as a whole. The predictive power of these groups, or functional masks, are given a fitness measure based on their relative Shannon entropy and the frequency with which observations occur.

This functional analysis procedure is performed over the simulated trajectories; the functional mask discovered is used to perform growth prediction. The functional mask discovered is used as the basis for probabilistic parameter estimation and prediction of tumor growth. In this fashion, growth predictions are made with significant accuracy in a very general fashion; no domain knowledge about tumor growth or the model being used is incorporated in the analysis and prediction procedure.

Robert Coop is a PhD student at the University of Tennessee in the Machine Intelligence Laboratory. His research interests include machine learning, artificial intelligence algorithms, genetic algorithms, and discrete event system simulation. His work is primarily focused on machine learning methodologies as applied to abstract non-linear domains, and other applications of adaptive optimization methods to problem domains with high dimensionality. Robert Coop holds a B.S. in computer science and a M.S. in electrical engineering from the University of Tennessee and is a student member of the IEEE.

James Nutaro is a member of the Modeling and Simulation Group in the Computational Sciences and Engineering Division at Oak Ridge National Laboratory. His research interests include discrete event and hybrid systems, parallel discrete event simulation, modeling methodologies, and event based numerical methods. This work covers a broad range of application domains, with particular emphasis on control systems operating over IP-based communication links, communication and control in electric power systems, simulating of large transportation systems, and modeling and simulation of wireless communication networks. James Nutaro has authored or coauthored over 25 papers in the open literature. He holds a B.S., M.S., and Ph.D. in computer engineering from the University of Arizona and is a member of the IEEE.

%d bloggers like this: