2009 Spring
January 29, 2009
Justin Newberg
PhD student, Dept. of BME, Carnegie Mellon University
Automated Analysis of Subcellular Protein Patterns in Human Tissues
Systematic information on the subcellular distributions of proteins is required for more accurate cell models that can be applied to clinically relevant cases; moreover, such information plays an increasingly important role in medical diagnoses. Given the number of proteins, conditions, and cell and tissue types for which information is needed, there is a critical need for automated, high throughput acquisition and analysis of subcellular location patterns. Automated pattern recognition methods have been shown to be effective at determining protein patterns in limited cell culture datasets. We have adapted these machine learning methods to analyze protein patterns in tissue images. For this purpose, we have used the extensive collection of images in the Human Protein Atlas, which contains over 6000 proteins in immunohistochemically stained tissues. Our initial work on a subset of the Atlas showed that we can determine protein locations across 45 different tissue types with a high degree of accuracy. In this talk, I will discuss various methods- such as classification, clustering, and segmentation- for scaling automated analysis to a larger set of proteins in the Atlas, and I will show preliminary results obtained using these methods.
2008 Fall
December 4, 2008
Charles Jackson
PhD student, Dept. of BME, Carnegie Mellon University
Intelligent Acquisition and Model Building of Fluorescence Microscope Time Series
Fluorescence microscopy is a powerful tool for live cell imaging. However, it suffers from limited spatial and temporal resolution, and photobleaching and phototoxicity limit the duration of acquisition. Furthermore, with an explosion in the amount of data being acquired, visual inspection becomes impractical. In this talk, I will discuss an active learning framework that uses an intelligent acquisition system to efficiently obtain information from a time series. The intelligent acquisition system chooses which pixels to acquire, at which time points to acquire, and when to stop acquisition. I will show some preliminary results obtained from applying this framework to the 3T3 dataset provided by MurphyLab.
November 13, 2008
Jelena Kovacevic
Professor, Depts. of BME and ECE, Carnegie Mellon University
Director, CBI
Active Mask Segmentation of Fluorescence Microscope Data Sets
I will talk about the new active mask (AM) framework and an algorithm for segmentation of digital images, particularly those of punctate patterns from fluorescence microscopy. The AM segmentation framework is suited for digital images. It is based on a local majority voting-based scheme, and can incorporate different forms of the voting function as well as several different functions to skew the voting to obtain a meaningful segmentation result. This framework has multiresolution and multiscale techniques built into it and can be instantiated to segment data of any dimension. We demonstrate the efficacy of the AM through an algorithm for segmenting punctate patterns of cells in fluorescence microscope images. While the theory opens up interesting vistas for research and development, the results demonstrate AM's utility in practice.
October 2, 2008
Chakra Chennubhotla
Dept. of Computational Biology, School of Medicine, University of Pittsburgh
Spectral Methods for Multi-Scale Feature Extraction and Data Clustering
We present a new framework for feature extraction and dimensionality reduction, called Sparse Principal Component Analysis (S-PCA). In this algorithm, we introduce a sparsity constraint on the elements of an orthonormal bases matrix. We show how this constraint can help recover object-specific structure in a low-dimensional subspace in a local, scale-dependent form. The learning algorithm of S-PCA is very simple, consisting of successive planar rotations of pairs of basis vectors. The principal advantages of S-PCA over a standard PCA-based representation include an intuitive understanding of the features underlying the high-dimensional data ensemble and efficiency in computations resulting from a sparse basis representation. We will explore both these themes in the talk. Additionally, using S-PCA we present a new approach to the problem of contrast-invariant pattern detection. The novel contribution of this work is the design of a perceptual distortion measure for image similarity, i.e., comparing the appearance of an object to its reconstruction from the principal subspace.
The other issue that is fundamental to the analysis of naturally occurring datasets is how to cluster items in a dataset using pairwise similarities between the elements. To this end we present a spectral method called EigenCuts. Using a Markov chain perspective, we characterize the spectral properties of the matrix of transition probabilities, from which we derive eigenflows along with their half-lives. An eigenflow describes the flow of probability mass due to the Markov chain, and it is characterized by its eigenvalue, or equivalently, by the half-life of its decay as the Markov chain is iterated. A ideal stable cluster is one with zero eigenflow and infinite half-life. The key insight in this work is that bottlenecks between weakly coupled clusters can be identified by computing the sensitivity of the eigenflow's half-life to variations in the edge weights. The EigenCuts algorithm performs clustering by removing these identified bottlenecks in an iterative fashion. As an efficient step in this process we also propose a specialized hierarchical eigensolver suitable for large stochastic matrices.
2008 Spring
April 3, 2008
Tim W. Nattkemper
University Bielefeld, Faculty of Technology, Biodata Mining & Applied Neuroinformatics Group
Information systems for biodata
Nowadays the volume and dimensions of biological data increases rapidly which calls for new information systems that support users in many data analysis tasks. The research of the group Applied Neuroinformatics & Biodata Mining aims at the development of new techniques to support users in data mining, visualization, exploration and retrieval. A special focus lies on the application of learning algorithms and data-driven approaches. In my talk I will give a small overview on our projects including multivariate bioimage analysis, metagenome data visualization and large scale underwater image evaluation.
March 26, 2008
Kerem Pekkan
Assistant Professor, Dept. of BME, Carnegie Mellon University
Imaging Challenges and Opportunities in Cardiovascular Fluid Dynamics
February 7, 2008
Sumit K. Nath
RPI
An Algorithm for Computing Accurate Neighborhood Relationships Using GVD
I will present an algorithm for computing accurate neighborhood relationships between arbitrarily-shaped objects using the Generalized Voronoi diagram (GVD). This algorithm has been widely studied in the context of path planning and terrain modeling. Assuming, we have apriori knowledge of the objects in a scene, then, using principles of Hamilton-Jacobi skeletons it is easy to compute the GVD. From a biological perspective this algorithm is useful for reliably segmenting and tracking cells without imposing constraints (e.g., the cell count remains constant during tracking). I will compare the performance of this algorithm with another recent algorithm developed by Kalra, et at CMU as well as traditional label propagation algorithms like the Watershed algorithm.
January 28, 2008
Rami S. Mangoubi
C.S. Draper Laboratory, Cambridge, Massachussetts, USA
Simultaneous Smoothing and Segmentation for Brain Imaging
Variational simultaneous smoothing and segmentation is a powerful method that carries several advantages. For instance, it can handle nonlinear objectives and non-Gaussian noise. It is also generalizable and can incorporate constraints. Brain imaging is one application area that benefited from this flexibility. The talk will survey a few problems in brain imaging that we have addressed with this method: functional MRI, diffusion tensor imaging (DWI), and angiography. Each of these problems motivates original contributions to the variational approach. We will discuss the challenges that each of these applications present, then choose one of these applications and show how our formulation provides improved results.
2007 Fall
November 14, 2007
Edwin Lughofer, Bettina Heise
Fuzzy Logic Laboratorium Linz-Hagenberg, Austria
An On-Line Interactive and Self-Adaptive Image Classification Framework
In many machine vision applications, such as inspection tasks for quality control, an automatic system tries to reproduce human cognitive abilities. The most efficient and flexible way to achieve this, is to learn the task from a human expert. This training process involves object recognition methods, adaptive feature extraction algorithms and evolving classifiers. A lot of research has been done on each of these topics, however, simply plugging all of these methods together does not necessarily lead to a working machine vision system. In this talk, a generic self-adaptive image classification framework is presented, focusing on integration issues and on topics that are specific to quality control applications. The basic components of these framework are the following:
- Generating contrast images by calculating the deviation images to the master
- Recognizing ROIs in the deviation images
- An adaptive object and aggregated feature extraction component (object features characterize single objects, whereas aggregated features characterize whole images)
- Training of (initial) base classifiers for aggregated and object features (aggregated and object classifiers) based on off-line pre-labeled image sets
- Ensemble Classifiers for resolving contradictory input among different operators
- Strategy for on-line classification of new images (incorporating object, aggregated and ensemble classifiers)
- On-line adaptation/evolution of base and ensemble classifiers (based on operator’s feedback during on-line mode)
- Early prediction of success or failure of a classifier
October 30, 2007
Justin Crowley
Assistant Professor, Dept. of Biological Sciences, Carnegie Mellon University
Toward an automated analysis of neural circuitry
Systems neuroscience is the study of the structure and function of complex neural circuits. A key set of tools employed by the systems neuroscientist are neuroanatomical tracing methods. These reveal the connectivity patterns of the brain - its wiring diagram - but require either extensive human labor for data collection and analysis or the development of new, automated image acquisition approaches. I will introduce neural tracing tools for systems neuroscience and discuss some structure-function relationships in the the visual system explored by my research group. In addition, I will describe some of the neuroanatomical data acquisition and analysis tools available currently and my wish list for future development in the field.
September 27, 2007
Kang Li
Ph.D. student, Dept. of ECE, Carnegie Mellon University
Cell Population Tracking and Lineage Construction with Spatiotemporal Context
Automated visual-tracking of cell populations in vitro using phase contrast time-lapse microscopy is vital for quantitative, systematic and high-throughput measurements of cell behaviors. These measurements include the spatiotemporal quantification of migration, mitosis, apoptosis, and cell lineage. The low signal-to-noise ratio of phase contrast microscopy images, high and varying densities of the cell cultures, topological complexities of cell shapes, and the wide range of cell behaviors pose many challenges to existing tracking techniques. We present an automated cell tracking system that can simultaneously track and analyze thousands of cells. The system performs tracking by cycling through frame-by-frame track compilation and spatiotemporal track linking, combining the power of two tracking paradigms. We applied the system to a range of cell populations including adult stem cells. The system achieved tracking accuracies in the range of 85.9%-92.5%, outperforming previous work by up to 9%. The proposed tracking methodology is valuable for tissue engineering, stem cell research, drug discovery and development, and related areas.
2007 Spring
May 10, 2007
Amina Chebira
Ph.D. student, Dept. of BME, Carnegie Mellon University
Adaptive Multiresolution Frame Classification of Biological and Biometric Images
We propose to develop an adaptive multiresolution frame classification algorithm for classification of biological and biometric images. This is motivated by the need for automated, accurate and efficient systems to extract knowledge contained in such image data sets. While biological data sets are typically processed visually by biologists, their sheer volume and dimensionality---a direct result of the revolution of “omics” projects such as genomics and proteomics, as well as advances in biochemistry, probes, and microscopy---make visual inspection error-prone, nonreproducible, subjective, and finally, impractical. By looking into four different biological/biometric applications, we find that the underlying problem is classification and that an accurate and efficient classification algorithm would be of great use to biologists, motivating the developments in this work. An automated classification algorithm was used in the problem of recognizing proteins based on the images depicting their location within the cell. As in that attempt, the authors had success with the simplest of multiresolution techniques, we postulate that using more sophisticated ones would lead to more accurate classification. Nonredundant multiresolution tools---multiresolution bases, in their adaptive incarnation, have been used with great success in fingerprint recognition. In the same problem, the authors observed that the translation variance of these bases might pose a problem and suggested to consider redundant multiresolution techniques---frames. Having motivated the use of adaptive MR in classification as well as the need for redundantMR transforms, we test that hypothesis by developing an adaptive MR classification algorithm. Given its success in the four application domains we consider, we argue that new, application adapted, frame families are needed. We thus present our initial results on building a new class of frames we call lapped tight frame transforms (LTFTs), and proceed to show which gaps are left to fill leading to the proposed plan and our final goal: the development of an accurate and efficient adaptive multiresolution frame classification algorithm for classification of biological and biometric images.
April 19, 2007
David Tolliver
Robotics Institute, Carnegie Mellon University
Image Processing with Spectral Graph Methods
I'll discuss Spectral Graph Algorithms and their application to Image Processing problems. In particular a grounded explanation of the common artifacts induced by spectral methods will be explained and addressed by two approaches: Spectral Rounding and expansion augmented graph topologies. I'll show that these modified algorithms compare favorably to current spectral approximations on the Normalized Cut criterion (NCut). Results are given for natural image segmentation, medical image segmentation, and clustering. This is joint work with Gary Miller in the Computer Science Department at CMU.
March 22, 2007
Gowri Srinivasa
Ph.D. student, Dept. of BME, Carnegie Mellon University
Multiscale Active Contour Transforms for the segmentation of fluorescence microscope images
In recent years, the focus in biological science has shifted to understanding complex systems at the cellular and molecular levels, a task greatly facilitated by fluorescence microscopy. Segmentation, a fundamental yet hard problem, is often the first processing step following acquisition. Our team demonstrated that a stochastic active contour based algorithm together with the concept of topology preservation (TPSTACS) successfully segments single cells from multicell images. It is a framework that is amenable to modifictions to be applied to successfully segment images from other imaging modalities such as the DIC and MRI. While this method is a viable alternative to hand-segmentation, it is not ready to be used for high throughput applications due to its large run time. In our present work, we highlight some of the benefits of combining TPSTACS with the multiresolution approach for the segmentation of fluorescence microscope images. While ideas along these lines have appeared in literature, there is no framework that allows us to elegantly develop new algorithms to accommodate for the changes in data (cell line, experimental conditions etc.) or even extract features from the fluorescence microscope images, hitherto unused, that aid in their segmentation. These motivate us to revisit the mathematics. Herein we propose a multiscale active contour transformation paradigm that provides a powerful mathematical framework for developing a family of flexible and modular algorithms for the segmentation of fluorescence microscope images in particular and biomedical images in general.
|