Semester

Fall

Date of Graduation

2006

Document Type

Dissertation

Degree Type

PhD

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Donald A. Adjeroh.

Abstract

To study the development of nervous systems, biologists are interested in neurite outgrowth, differentiation, synapse formation, and plasticity. High throughput neuron image processing is an important method for drug screening and quantitative neurobiological studies. The power of high-throughput processing comes from the automated fluorescence microscopy imaging techniques that make it possible and facile to visualize the complicated biological processes on the cellular and molecular levels and allow fast and cheap acquisition of such imaging data. With this method, a huge number of images are generated, with resolutions at length scales that are small enough to resolve neuron structures. Thus, one immediate challenge facing researchers now is to find efficient and effective methods for managing the unprecedented volume of image data. Accessing these data to generate useful knowledge requires efficient and effective image analysis tools that involve the smallest human interaction. In our work, we study two problems related to neuron images.;The first problem is on lossless compression of neuron images. We consider context based modeling methods, which are seen an important step in high performance lossless data compression. We study methods for effective context modeling for images based on existing successful modeling methods used for text compression. A novel context based modeling method is proposed that is used to compress neuron images in a lossless manner. We also extend the modeling method to compressing other types of images, including natural images.;The second problem is on neuron structure extraction from neuron images. The neuron structures, including curvilinear neurite segments and dendritic spines, exhibit the connectivity of the neural networks and thus can be used to study the functionality of the neural networks. The extraction and analysis of the neuron structures are still accomplished manually, or semi-automatically. Thus, we are interested in developing fast and fully automatic algorithms for extracting neuron structures. For this purpose, we develop novel methods for extracting curvilinear neurite segments in 2D neuron images and for extracting dendritic spines in 3D neuron images. We also study effective validation methods for evaluating the performance of the proposed neuron structure extraction algorithms.

Share

COinS