Center for Statistical Sciences
Department of Radiology, Brigham and Women's Hospital
Department of Health Care Policy, Harvard Medical School
Abstract: The validity of image segmentation is an important issue in image processing because it has a direct impact on surgical planning. We examined classification accuracy in imaging analysis based on three two-sample validation metrics against the estimated composite latent gold standard, which was derived from several experts' manual segmentations by an expectation-maximization (EM) algorithm called simultaneous truth and performance level estimation (STAPLE). The distribution functions of the tumor and control pixel data were parametrically assumed to be a mixture of two beta distributions with different shape parameters. We estimated the corresponding receiver operating characteristic (ROC) curve. Dice similarity coefficient, and mutual information, over all possible decision thresholds. Based on each validation metric, an optimal threshold was then computed via maximization. We illustrated these methods using magnetic resonance (MR) imaging data on three radiological examples: (1) accuracy of brain tumor segmentation, (2) functional imaging for detection of brain activation, and (3) hidden gold standard in prostate peripheral zone segmentation for brachytherapy. Extensions of incorporating spatial correlation structures were also considered under a Markov random fields model. al Key Words: Sensitivity, Specificity, Receiver operating characteristic (ROC) curve, Dice similarity coefficient, Mutual information, Expectation maximization (EM) algorithm, magnetic resonance (MR) imaging, Functional MR (fMRI).
*Center for Computational Molecular Biology Seminar Series*
Biomedical Engineering Department, Cambridge, MA
Abstract: Recent analyses of high-throughput protein interaction data coupled with large-scale investigations of evolutionary properties of interaction networks have left some unanswered questions. To what extent do protein interactions act as constraints during evolution of the protein sequence? How does the type of interaction, specifically transient or obligate, play into these constraints?
Are the mutations in the binding site of an interacting protein correlated with mutations in the binding site of its partner? We address these and other questions by relying on a carefully curated dataset of protein complex structures. Results point to the importance of distinguishing between transient and obligate interactions.
We conclude that residues in the interfaces of obligate complexes tend to evolve at a relatively slower rate, allowing them to coevolve with their interacting partners. In contrast, the plasticity inherent in transient interactions leads to an increased rate of substitution for the interface residues and leaves little or no evidence of correlated mutations across the interface.
Brown Analysis Seminar
Scientific Computing Seminar
Abstract: The evolution of cosmic baryon fluid is dynamically governed by the underlying dark matter field, as the cosmic gravitational field is dominated by dark matter and dark energy. In linear reqime, the velocity and density fields of baryon matter follow the dark matter point-by-point. However, once the nonlinear and stochastic nature of this system should be considered, the dynamical behavior of baryon fluid is significantly different from those of the dark matter, and shows turbulence-like features, such as intermittency, abnormal scaling, statistical decoupling. I will present the observational evidences and theoretical reason of these features. To understand this evolution, high quality computer simulation of cosmological hydrodynamics is critical. Some applications of the WENO/N-body code and its ongoing problems in this field will also be addressed.
Student Pizza Seminar
THIS TALK HAS BEEN CANCELLED FOR TODAY DUE TO ILLNESS
Department of Mathematics Colloquium Lecture
<--- 2005 Index
DAM Home Page