Bayesian network modeling of neural systems with functional MR images

• Main Author: Zheng, Xuebin
• Other Authors: Rajapakse, Jagath Chandana
• Format: Theses and Dissertations
• Published: 2008
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence; DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision; DRNTU::Engineering::Computer science and engineering::Computer applications::Life and medical sciences
• Online Access: https://hdl.handle.net/10356/2585
• Institution: Nanyang Technological University

Neuroscientists have shown increased interest in knowing interactions among brain regions activated during sensory and cognitive tasks. The existing methods of connectivity analysis are confirmatory in the sense that they need a prior connectivity model to begin with. These methods are often under anatomical constraints or complicated by the fact that many of the prior model have been obtained in the studies of monkeys. This thesis presents an exploratory (data-driven) approach based on Bayesian networks in modeling neural systems with functional MR images. Bayesian networks are directed graphs where the effective connectivity between two brain regions is represented by conditional probability densities (CPD). Therefore, the interactions in the network are represented in a complete statistical sense. The previous methods of testing disconnectivity hypotheses in brain diseases were mostly done by comparing the activation patterns. However, this is not effective when the diseases are due to deficits in connectivity. This thesis demonstrates how the graphical models derived by the present method can be effectively used in the analysis of lesion studies. These studies indicate the promise of the present method as a general framework for analyzing a wide range of brain disorders in future.