Research In Bioinformatics

The Bioinformatics lab at CIDDS provides computational facilities and bioinformatics support for the researchers in various colleges and departments at Ohio University. The current research covers the area of Functional Genomics, Proteomics and Motif Discovery. Our activities include the development of a platform for managing in silico experiments, administering and distributing research data, developing tools to analyze those data, and providing the expertise to continue data intensive biological research. The Bioinformatics lab is located in 304 Stocker Center.

Research Projects

Project: Proteomics Resarch at Edison Biotechnical Institute

CIDDS is assisting Edison Biotechnical Institute (EBI) with functional proteomics research.  CIDDS is analyzing the process of functional proteomics research and providing custom integrated software to expedite the research.  The current version of the software increases research potential by automating many of the difficult repetitive tasks performed by lab technicians.  In addition, the software features new research methods and algorithms for protein identification.  These new processes promise to advance this field of Proteomics Research at Ohio University.

Project:  Functional Genomics Research at Environmental and Plant Biology

CIDDS is working closely with Plant Biologists conducting research into the function of plant cell wall proteins.  CIDDS has incorporated numerous custom algorithms into a computer program that can analyze vast amounts of genome data and produce summary results very quickly.  This software advances the field of Functional Genomics Research by allowing the Biologist of Bioinformatician to design, create and execute new ad hoc experiments with ease.  The current version of the software is being used to reproduce previous studies, automate everyday repetitive tasks, and test new hypotheses without the need for a lot of tedious bench testing.

Project:  Gene Expression and Motif Discovery Research at Environmental and Plant Biology

CIDDS is working closely with Plant Biologists conducting research into the Motifs and Mechanisms of gene expression in plants.  This field of research is very computationally intensive.  Bioinformaticians at CIDDS have created an analysis tool comprised of existing algorithms and research into new algorithms.  This software automatically analyzes complex genome data using several algorithms simultaneously.  This software advances research in the field of Gene Expression and Motif Discovery by allowing researchers to quickly analyze the relative merits of previous research methods as well as new custom or hybrid approaches.

Project: Registration Assisted Image Smoothing and Segmentation

Registration and segmentation are two most important problems in the field of medical image analysis. Traditionally, they were treated as separate problems. In recent years, researchers started to consider exploiting the dependence exisiting in these two problems to achieve better performance for each of them. In this project, we investigate a segmentation-guided registration framework which integrates the available segmentation information into the process of registration. Unlike other models, our model can very naturally handle not-rigid as well as rigid deformation. Comparing with the corresponding intensity-only models, our segmentation + registration produces a more stable and noise-tolerant mapping estimation process, where regions of interests (ROIs) are better captured.

Project: Robust Multimodal Image Registration Based on Artifacts Reduction

Technological advances in medical imaging in the past two decades have presented many medical disciplines with large quantities of images, and it is becoming more common to fuse these data sets with image guidance systems to aid in the application of therapeutic procedures.

Fusing of multi-modal data involves automatically estimating the coordinate transformation required to align the multi-modal image data sets. Mutual information (MI) is currently the most popular match metric in handling the registration problem for multi modality images. However, interpolation artifacts impose deteriorating effects to the accuracy and robustness of MI-based methods. We analyze the generation mechanism of the artifacts inherent in the interpolation procedure and prove that uniform interpolation kernels are the major cause of registration artifacts. As a remedy, we preopose new registration algorithms based on non-uniform interpolators. Preliminary data already demonstrate the great potential of these alignment algorithms in improving the robustness and accuracy of mutual information. A multi-thread implementation incorporated in a user-friendly interface will be developed.

Project: Human Motion Tracking for Automated Scene Understanding

For a typical CCTV-based scene understanding system, the functional structure can usually be divided into four serial parts:

• initialization
• tracking
• pose estimation
• recognition

Tracking, as the step to provide the inputs for later pose estimation and recognition processings, determines the overall performance of a system, which makes it an important problem in CCTV video system. Due to the non-rigid nature of objects that being tracked, deformable models are being widely used in tracking tasks.

The proposed project will signifcantly improve the reliability of the surveillance and ground reconnaissance video by providing a robust and efficent object tracking framework. In particular, this research will: 1) provide a platform for integration and demonstration of motion segmentation, robust real-time tracking, recognition and adaptive compression and indexing; and 2) experiment a new robust tracking approach as well as a search interface based on the occurrence of target objects in the video, with the goal to identify all spatial/temporal related objects.

Recent publications

1.  Jundong Liu, Chunming Liu and D. Fox, Segmentation of external force field:  An approach toward automatic initialization and splitting of snakes, Pattern Recognition, Vol. 38 (11), Nov. 2005, pp. 1947-1960.

2.  Libo Cao, Peter B. Harrington and Jundong Liu, SIMPLISMA and ALS Applied to Two-dimensional Nonlinear Wavelet Compressed Ion Mobility Spectra of Chemical Warfare Agent Simulants, Analytic Chemistry, 77:8 (2005) pp. 2575-2586.

3.  Jundong Liu, Vector-valued Local Frequency Representation for Robust Multimodal Image Registration, accepted to 27th Annual International Conference of EMBS (EMBS'05), Shanghai, 2005.

4.  Jundong Liu and Yang Wang, Segmentation Guided Robust Multimodal Image Registration using Local Correlation, accepted to 27th Annual International Conference of EMBS (EMBS'05), Shanghai, 2005.

5.  Jungdong Liu, Segmentation Guided Registration for Robust Brain Atlas Construction, Human Brain Mapping, Toronto, Ontario, June 12-16, 2005.

6.  Chunming Li, Jundong Liu, and Martin D. Fox,  "Segmentation of Edge Preserving Gradient Vector Flow:  An Approach Toward Automatically Initializing and Splitting of Snakes,"  IEEE Conference on Computer Vision and Pattern Recognition (CVPR'05), San Diego, June 22-24, 2005.

Schedule for Weekly Research Meetings