Research Activities

Past Research Projects

Interactive Detection, Visualization and Quantification of Breast Lesions from DCE-MRI Volumes

Mammography is currently regarded as the most effective and widely used method for early detection of breast cancer, but recently its lower sensitivity in certain high risk cases has been less than desired. The use of Dynamic Contrast Enhanced MRI (DCE-MRI) has gained considerable attention in recent years, especially for high risk cases, for smaller multi-focal lesions, or very sparsely distributed lesions. This project is focused on constructing interactive visualization tools to identify, process, visualize and quantify lesions from DCE-MRI volumes. Our approach segments the time varying volume data in near-real time and employs 3D texture mapped volume rendering in a highly interactive environment. (Joint Work with Dr. John Brockway, Novant HealthCare, Charlotte, NC).

Sample 3D Reconstructions:

• Invasive Ductal Cancer
• Benign Fibroadenoma
• Ductal Carcinoma In Situ (DCIS)
• Ductal Carcinoma In Situ (DCIS)

Modeling and Visualization of DNA Microarray Data

Whole genome expression profiling using DNA microarrays represent a major advance in genome-wide functional analysis, for understanding gene function, gene interactions, metabolic pathways, signal transduction pathways, and effects of environmental factors. The central challenge is to rapidly and efficiently extract useful information and discover knowledge from these large data sets, which can improve our comprehension of human disease, and facilitate drug development. This project employs information visualization techniques to view and interact with the output of modeling techniques (such as Graphical Gaussian, Loglinear, Causal) applied to DNA Microarray data. The output is in the form of multivariate relationships. The complexity and nature of these relationships motivates the use of visualization techniques and is expected to lead to new knowledge of biological pathways.

3D Reconstruction and Visualization of Intravascular Ultrasound Images

This project involves the reconstruction and visualization of 3D Intravascular Ultrasound image sequences, study and characterization of cholesterol induced blockages (atherosclerotic plaques). In the first part of this study, two vessels (bovine and dog arteries) were imaged, in vitro, reconstructed in 3D and visualized. The reconstructions were validated both qualitatively and quantitatively and an error within 10% was achieved. Future work on this project will involve 3D system development and in-vivo studies (Sponsored by the American Heart Association. Joint work with Heineman Medical Research Laboratory, Carolinas Medical Center).

Sample Intravascular Ultrasound Reconstructions

Animation (MPEG, 3.8M Bytes)

Interactive Visualization of Mobile Network Simulations

This project explores information visualization schemes to interactively control, drive and analyze simulations of adaptive resource management protocols in mobile networks. In particular, the research will use interactive steering to optimize multi-layer adaptive protocols, via visualization interactions of appropriate simulation variables (system monitoring metrics, metric parameters and simulation system parameters)(Joint work with Dr. Teresa Dahlberg, UNC Charlotte)
(Sponsor: National Science Foundation).

Medical Volume Modeling/Representation Using Variational Implicits

This project explores new techniques for active contour models, popularly known as snakes, by combining active segmentation models with constrained implicit surface models recently reintroduced under the name of variational implicit surfaces. Ongoing work explores algorithm for efficient representation of large medical models using variational implicits, computationally more efficient basis functions, etc.

Example CT Reconstruction of a Tooth Using Implicit Snakes
Example Intravascular Ultrasound Reconstruction of a Blood Vessel

Display and Analysis of Spinal Motor Circuit Simulations

This project will provide a visual display of output data and a rapid data-analysis package for simulations of spinal cord circuits. The particular circuits of interest are those controlling hindlimb muscles of a mammal, such as cat or human. The data, consisting of time and identity(location) of active cells in each population at each millisecond, are generated by a large-scale computational model of a pair of motoneuronal populations interconnected with ten or more interneuronal populations. The two motoneuronal populations represent the output to a pair of antagonistic (i.e., flexor-extensor) muscles. Neuron electrical properties and connection patterns are reasonable representations from the literature. The network contains on the order of 2500 cells and 650,000 connections. Inputs to the circuit consist of a number of fiber populations representing sensory inputs and descending control from the brain.

The visualization package will provide display of time and location of active cells for all the populations, as well as computing elementary cell and population performance statistics. An important goal of the project is to combine the simulations (experiments) and the data analysis (which will also be, for the most part, graphical) to function in a single computing environment. This is expected to drastically reduce the turn-around time from experiment to analyzed data (eight fold reduction is estimated). (In collaboration with Dr.David Bashor, Dept. of Biology)

The sample animation below shows the activity of three state variables, membrane potential (green), threshold potential (yellow) and spike production (grey) over a time interval of 66 milliseconds. The six populations on the left and the the lone population on the upper right corner are fiber populations, while the remaining 15 are cell populations. Each population represents 100 cells (neurons), arranged within a 10 by 10 grid.
(In Collaboration with Dr. David Bashor, Dept. of Biology, UNC Charlotte)

Sample Image(JPEG)
Sample Quicktime Animation (6 Mbytes)

Interactive Visualization of Medical Images Using Partitioning Trees

The goal of this project is to convert medical images (CT, MRI, etc) from a discrete lattice of values into a partitioning tree representation that is piece-wise continuous. This enables many viewing and spatial operations to be performed with great ease and at interactive rates. Applications include interactive visualization of medical data, surgery planning/rehearsal/training and prostheses design (Joint work with Dr. Bruce Naylor, University of Texas at Austin).

Synthetic Image: Lenna

Medical Image : MRI Brain Images

MS/Phd Theses Topics

IDS Data IDS Data (TAR FILE)