CS237 Project Idea List

These ideas, most less-than-completely-formed-and-described, were suggested by various researchers around campus. If you are interested in following them up for your project, please contact the provider of the idea or chat with me (dhl) for more information.

From David Laidlaw (dhl@cs) Computer Science

I'll describe a number of possible areas from which projects could be defined. In almost all of the areas below, initial steps have been taken to get needed data and to scope out a larger research agenda.

Evaluate, via a user study, how various kinds of textures can be used to encode multi-valued data.

Develop anatomical visualization for brain tumor pre-surgical planning.

Evaluate whether visualizing neural diffusion rate information is useful.

Defining and testing new DTI (Difussion Tensor Imaging) metrics in collaboration with medical researchers from Brown's hospitals.

Represent uncertainty visually.

We also have a number of grants that have recently been funded. Ask David about them if you are interested.

NSF(ITR): Art and perception for scientific visualization
NSF(ITR): Understanding Unsteady Bioflows through Simulation, Modeling, Visualization, Art, and Psychology (See calendar page for a link to the proposal)
NIH: DTI+MRI-based Tools for Analyzing White Matter Variation (See calendar page for a link to the proposal)
NSF(ITR): Collaborative Research: Perceptual Optimization for Data Visualization. This proposal is aimed at helping to establish a discipline of information psychophysics. The essence of information psychophysics is to apply the methods of classical psychophysics to the perception of common information structures, such as elementary flow patterns, surface shapes or paths in graphs. Much of human intelligence can be broadly characterized as the ability to identify patterns, and the visual system is the most sophisticated pattern finding mechanism that we have. Of all of our perceptual systems, vision dominates. It is estimated to engage 50% of the cortex and 70% of all our sensory receptors are visual, but is only just becoming possible to display as much information as the human visual system is capable of absorbing.
NSF(DDDAS): Interactive Data-driven Flow-simulation Parameter Refinement for Understanding the Evolution of Bat Flight. We propose a four year multi-disciplinary research project to create a Dynamic Data Driven Application aimed at improving the understanding of a complex biophysical system--the flight of bats. The system is comprised of a multi-level hierarchical simulation of bat flight based on parameterized features of bat morphology and behavior. The simulation is capable of very rapid solutions but requires input from measurements to ensure fidelity and optimality. This input will be provided in an integrated fashion, drawing from a parallel series of experiments in which several discrete data streams will be generated, including kinematic wing data, wave velocity data over a series of two dimensional planes, as well as other data such as bone deformation, experimentally determined material properties, etc. This data will direct the simulation ensuring accurate solutions, but still with high responsiveness. The data streams will be monitored, synthesized, combined and processed using an advanced immersive visualization environment which will be used to guide the interactions between the measurement and simulation and to organize the disparate streams of data.
NIH: 3D Multi-Articular models of the Carpus. Musculoskeletal disease and trauma of the wrist affect all age groups and account for significant medical expenses and lost earnings each year. The skeletal base of the wrist is the carpus consisting of eight small and anatomically complex carpal bones. Because there are no substantial muscular attachments on the carpus, the stability, use and pain free function of the wrist is wholly dependent upon the multiple cartilaginous articulations and the ligamentous structures of the carpal bones. While the mechanics of the wrist have been the focus of much previous work, a complete understanding of the mechanics of the carpus is lacking. Previously we have developed methods to study the in vivo kinematics of the carpal bones and have found these to be more complicated while finding unique patterns associated with the dart thrower.s wrist motion. Our previous studies were limited to the kinematics of the carpal bones with no information on cartilage or ligaments because of our use of CT imaging. We propose to develop a validated model based upon in vitro kinematic data, clinical CT and micro-CT volume images that will permit the study of normal deformations of the cartilage and ligaments. Current sophisticated models developed for the knee, for example, cannot be applied to the carpus because of the large number of carpal articulations and ligaments and the lack of digital data and mechanical properties for the carpal tissues. The proposed model would be a significant advance in biomedical computing because it will be capable of handling the carpus and other multi-articular structures, and it will incorporate complex multi-body kinematics. The model in this proposal will be limited because of the lack of detailed material properties, but the incorporation of the kinematics and digital anatomy will result in an innovative, sophisticated multi-articular model that will also provide the end-conditions for the deformable tissues - critical data for the development of finite element based models.

Gregory Jay MD-PhD (GJay@Lifespan.org)
Department of Emergency Medicine, RI Hospital
Division of Engineering, Brown University

Studies of diarthrodial joint lubrication have been conducted over the last 70 years. A mechanism employing boundary lubrication has been postulated to explain how articulating cartilage surfaces under load, lubricated by synovial fluid, produce very low coefficients of friction (~ 0.01) or less. Even our best manufactured load-bearing systems, lubricated by Teflon, produce significantly higher friction (~ 0.04). A molecule called lubricin in present in synovial fluid which endows this fluid with these special properties. Boundary lubricants in general must be capable of binding to a surface and generate repulsion by some physical means. The O-linked alpha(2,3)NeuAc-beta(1,3)Gal-GalNAc in the central mucin-like domain contributes to the protein’s boundary lubrication of the cartilage surface possibly due to repulsive hydration forces or charge repulsion. Recently preliminary molecular models reveal how lubricin has surfactant-like activity. There is an opportunity to build on these models which would help us understand how the lubricating layer on cartilage deforms and resists shear.

Matthew Harrison (mth@dam.brown.edu)
Post-Doc at the Applied Mathematics Dept.
Brown University

We are interested in understanding the stimulus-response functions of low-level visual neurons. Mathematically, this is idealized as a (usually nonlinear) function of a small-medium image patch (or perhaps a temporal sequence of such patches). Given such a function for a neuron (either hypothesized or fit from actual neural data), we want to get an idea about "what the neuron is looking for" and "what it is invariant to". That is, what sorts of image patches maximize the function and what sorts changes to the image patch leave the response about the same. Both are easy to formalize mathematically, but the latter, especially, is difficult to visualize. Current approaches involve making little movie snipets of a deforming image, all of which give the same response. This is not really great. It would be much better to interactively explore the landscape of this response function and really get a feel for what is going on.

Kebring Yu (contact Arthur_Salomon@brown.edu )
Brown University

He measures activation of proteins as part of the signaling processes that underly the
operation of cells. The experimental results that they gather can be
used together with data from other publicly available sources to
reason about how sequences of signaling processes work. A tool that
helped to visually analyze these different kinds of data could help
them reason more quickly and effectively.

Arthur R. Salomon (Arthur_Salomon@brown.edu )
Arthur Salomon Proteomics Lab
Bio Med Molecular, Cellular Biology Biochemistry
Brown University

The objectives of the group are to exploit emerging proteomic technologies to gain insights into cellular signaling networks important in understanding and treatment of disease. Secondly, we are interested in the development of bioinformatics tools to gain insights from the tremendous amount of data generated in modern protemics experiments.

Jan Bruder (janbruder@hotmail.com)
Bio-Med
Brown University

I am researching the interactions of neurons with 3dimensional substrate
materials to better understand neuronal behavior in vivo. One of the main
obstacles in assessing the influence of a particular substrate is the
precise visualization and quantification of neurons in three dimensions over
time.

We basically need an application that can assemble stacks of images captured
with phase or fluorescence microscopy (could be confocal) into 3dimensional
fully rotatable views which can then animated to allow 3dimensional analysis
of timelapsing data.

Additionally, it would be great if the application could be used to quantify
parameters such as cell volume and surface area, number of cell processes,
number of nuclei, and average orientation of the major axis. A manual
variable threshholding function should allow for user-adjusted selection of
relevant brightness levels.

Your application could greatly facilitate our research. I would be thrilled
if you chose to take on this application -- even partial functionality could
enhance our results significantly.

Selim Suner MD, MS(Ssuner@Brown.edu)
Department of Emergency Medicine
Brown University Medical School

Visualization of vascular structures and hand (needle position) could facilitate instruction of medical students and residents in central venous access. This technique can also be used as an adjunct in patients who have difficult access to venous access. The venous structures could be visualized using continuous 2D ultrasound images obtained in different planes to facilitate constructing a 3D image. These images will also give information regarding soft tissue density and depth above the venous structures of interest. (Laerdal makes a 2D computer screen trainer for this using a haptic device for feedback; in my experience this has not been not a very useful tool and can be improved upon).

Warren Prell (Warren_Prell@Brown.edu)
Department of Geological Sciences
Brown University

Visualization of Narragansett Bay 0ceanography

Naragansett Bay is central to the economy and quality of life in Rhode Island.� As a coastal estuary, the oceanography of Narrgansett Bay is dependent on mixing with the ocean, the fresh water flow into the headwaters, the introduction of nutrients and pollutants into the bay, and the general climate.� Because of these multiple inputs, the oceanography changes on a wide range of time scales from tidal variations (~6 hours), to event-scale storm responses (days), to seasonal cycles and finally to inter-annual changes that reflect climate.� Measuring the changes in Narragansett Bay on all these time scales is almost impossible.� However, a new data set has become available that enables a new view of the Bay's oceanography.� Unfortunately (fortunately?) the data set is large and difficult to visualize.� Hence, the opportunity to do something new.

�� The data are collected by an undulating sensor package that measures the temperature, salinity, dissolved oxygen, and chlorophyll of the water column as a function of geography.� Approximately 60 cruises are available; each represents a circuit around the bay during a specific month.� Thus, for the whole Bay, information is available on the water column in terms of depth and location and (if multiple data sets are used) by time of season or year.� This is a truly unique and underutilized dataset.� Measurements are made each second and each cruise takes 7 to 8 hours, so the files are about 30,000 lines.�� See the following url for information about the data:
http://www.narrbay.org/d_projects/nushuttle/shuttletree.htm

What can be done with the data?� Many questions can be addressed by the data. Examples might be: How does the distribution of hypoxia (low dissolved oxygen, DO) evolve on a seasonal scale.� How to efficiently visualize and quantify the difference in low DO between different years? How does stratification (vertical density gradient) evolve on a spatial and seasonal basis and how does it control the development of low DO? How can the data be used to visualize mixing between ocean and fresh waters?

Brad Marston(bradmarston@mac.com)
Department of Physics
Brown University

How to best visualize inviscid two-dimensional fluid flow in a way that makes manifest the infinite number of conservation laws [conservation of the vorticity probability distribution]?

Sorin Istrail (sorin@cs.brown.edu)
Department of Computer Science
Brown University

Cloning Murray "Gold Standard" Resnick, MD:

Viz tools for automatic diagnosis for surgical pathology of cancer. These tools would help increase the diagnostic accuarcy and decrease the inter-observer variability when evaluating pathological information.

Bob Pelcovits (pelcovits@physics.brown.edu)
Department of Physics
Brown University

Visualizing smectic liquid crystals which are composed of fluid layers.
Here's what I wrote in my NSF proposal last year:
We are also developing visualization tools for the smectic phases. As in the nematic phase we will use cubic b-splines to smooth our data, but in this case we will consider the mass density rather than the orientation field of the molecules. However, following a suggestion of Laidlaw, we will use the molecular orientation to help us establish the continuity of the smectic layers. E.g., we will consider the molecules as pancakes oriented normal to the local director field in the smectic A phase. The orthonormal vectors that span the pancake can then be used to generate a continuous surface, in much the same way that the streamsurfaces are generated for the nematic phase. This procedure should generate smooth smectic layers (possibly after some experimentation to accommodate interdigitation of the layers) and allow us to readily see and distinguish edge and screw dislocations and probe their structure.

Christophe Benoist, MD, PhD (Christophe.Benoist@joslin.harvard.edu)
Joslin Diabetes Center
Harvard Medical School

Studies on autoimmunity explore the immunological mechanisms of diabetes, rheumatoid arthritis and APECED. Major questions tackled are what initiates these diseases, how is their progression regulated, and what are the final effector mechanisms. In addition, modern genetic and genomic approaches are used to identify disease-modifying genes in both human patients and mouse models, and the application of computational and bioinformatic strategies to these and other issues is beginning to be explored.

Jerrold Boxerman, MD, PhD (JBoxerman@lifespan.org)
Dept. of Diagnostic Imaging,
Neuroradiology, RI Hospital

Our group is interested in using advanced MR imaging techniques to better
characterize and diagnose brain tumors. One diagnostic challenge is
distinguishing a primary, infiltrative brain tumor (glioma) from a solitary
metastasis or tumefactive inflammatory lesion. Diffusion tensor imaging
(DTI) offers the potential to make this distinction based upon differences
in tensor characteristics within brain tissue surrounding the lesion. New
image processing and visualization techniques are required to aid in
accurate diagnosis.

John Janotti (jj@cs.brown.edu)
Department of Computer Science
Brown University

Website designers would like their sites to be easy to use. As "web applications" have become more complex, web site usability has moved from the already difficult problem of clean layout and organization to the full challenge offered by traditional application development. Traditional applications are best evaluated with costly user studies to assess how long various tasks take, the shortcuts that users use or avoid, etc. These user studies are, by necessity, small and intermittent. Web-based applications offer the opportunity to constantly monitor the behavior of every user.
The challenge is to develop data gathering tools and visualizations for the data that help the application designer understand how the application is being used. A purely passive approach would analyze existing logs. This approach might animate how individual users moved from page to page, how long average users stayed on various pages, how often users abandoned checkouts at various phases of completion, etc. A more active approach might instrument the application to report more data or even to present alternate interfaces to various users in order to compare aggregate behavior.

Peter Schultz (Peter_Schultz@brown.edu)
Department of Geological Sciences
Brown University

I am a Brown Prof in geology and a Co-Investigagtor on NASA's Deep Impact Mission, which you may have heard about. This mission involved hitting a comet and observing while flying by in a companion spacecraft.
My students and I would like to explore the possibility of looking at the comet we hit with some of the graphics tools developed at Brown. More specifically, we have a preliminary shape (mathematical) model of the comet and we would very much like to reconstruct the effects of sun angle and view angle as the spacecraft zoomed by. We are not particularly versed in this type of graphics but was hoping there might be someone there who could help us out. We also have a mathematical description of the debris coming off the impact (based on experiments)and would like to test some models (shadows, projections) for comparisons with what we saw.
This mission is still in the stages where the data have not been released. As a result, there is also some friendly competition among groups, including the visualization lab at Cornell. I would very much like to show what we at Brown can do, particularly because we have some unique data sets for comparison (and are now rushing to get results ready for publication).

John F. Hermance (John_Hermance@Brown.Edu)
Environmental Geophysics/Hydrology
Department of Geological Sciences
Brown University

Computer Visualization Projects in Hydrology, Remote Sensing and Geophysics

Water in Nature (A): Floods develop in rivers through the behavior of water in individual channels (through characteristic time constants and discrete storage capacities), and the compositing of these sub-elements through inflowing tributaries. How can this be best visualized in time and space in order to demonstrate the evolution of a "flood event"?

Water in Nature (B): A fundamental process in groundwater flow is the response of the watertable (and the fluid in the saturated zone beneath) to the impulse of water added or withdrawn from above. Mathematical equations can be used to simulate this behavior, but how can the interaction of the velocity fields and pressure gradients be best visualized to demonstrate the flow of water in a dynamic way?

Earth's Vegetation from Space: Each day satellites acquire a snapshot of the Earth's vegetation on a global scale at a resolution of (often better than) 10 km x 10 km. The best of these data (i.e., cloud-free) are distributed at 1 month (often more frequent) intervals, providing regional time series at 1 month (or better) samples over a period of record of many years. We've developed algorithms for the robust interpolation of these samples in time (to step over dats gaps from cloud cover. etc.). What is begging to be done is to tease from the data certain information on overall vegetative patterns in space and time as synoptic static images, as well as animated sequences (or movies).

Exploring the Earth's Subsurface Non-Invasively: One of the most effective means for "probing" the Earth's subsurface for groundwater, environmental and engineering studies is to generate a wave from a source (which could be seismic, acoustic, electromagnetic or radar) on the Earth's surface, and then interpret the behavior of waves as they propagate through its interior. Simple models can be used to predict the time/space behavior of spherical waveforms as they interact with geologic/hydrologic layers, but how best can one use computer visualization to extract the fundamental aspects of the phenomena in order to provide the viewer with insight into the interaction of various reflected and refracted phases?

Peter Richardson (Peter_Richardson@brown.edu)
Division of Engineering
Brown University

In 'Optics and Laser Technology' on line Aug 2005 (accessible via Josiah) there are some articles about use of color, especially in visualizing complex fluid motions:
- Fryer MJ., "Complementarity", doi:10.1016/j.optlastec.2005.06.003
- Kennear D, Atherton M, Collins M, et al, "Colour in visualisation for computational fluid mechanics", doi:10.1016/j.optlastec.2005.06.015
- Stuecke P, Egbers C., "Visualization of scavenging flow in the design of small two-stroke engines", doi:10.1016/j.optlastec.2005.06.036
- Carlomagno GM., "Colours in a complex fluid flow", doi:10.1016/j.optlastec.2005.06.016
Take any or all of these for conceptual content and examine color in representing unsteady flows, e.g. exploring a complementarity sequence with 90 degrees phase shifts in a cyclic flow - how well can the sequential image frames embed something of current and phase-shifted flows simultaneously in a fixed geometry? Does this help in spotting locations where flow reversal occurs during a cycle of a cyclic flow?

Steve Correia (SCorreia@Butler.org)
Butler Hospital

Steve Correia studies aging and uses diffusion MRI as part of his studies He has many new ideas

about how new visualization tools could help in quantifying brain changes. He can suggest more specific project
ideas along those lines, likely building on some of our local tools that are already in place.

David Tate (DTate1@Lifespan.org)
Immunology
Brown University

David Tate studies HIV and also uses diffusion MRI. He has ideas
about how new visualization tools could help in quantifying brain changes. He can suggest more specific project
ideas along those lines, likely building on some of our local tools that are already in place and that he is using.

Sunil Shaw (SShaw@WIHRI.org)
Asst. Prof. of Pediatrics
Women and Infants Hospital

I am a biomedical scientist, an immunologist, and I am particularly
interested in how cells move around. I use videomicroscopy and molecular
biology tricks to get information about cell signals and structural
molecules, and thus build a conceptual model.

In white blood cell transmigration, white blood cells squeeze through a
blood vessel wall. (This results in inflammation - part of the immune
response). White blood cells typically squeeze between vessel wall cells,
and the gap closes up thereafter. I am interested in exploring the
possibility that the cells 'talk' to one another by means of intracellular
calcium (Ca). When I visualized intracellular Ca in these cells, using
video microscopy, I saw lots of oscillations in Ca levels (periodic rises
and falls). I want to determine if these oscillations are
communicated/propagated between cells. Due to sampling limitations, this
may require some clever correlation and statistics to prove.

I am proposing to provide some primary data (movies of cells moving, with
their Ca levels). I would also probably need to talk with students to
explain some of the context.
Out of this, I am looking to get an answer, or at least an
analysis/visualization technique to whether these cells communicate.

Leslie Degroot (Leslie_Degroot@brown.edu)
Endocrinology
Brown University

A program for visualizing the fit of peptides, related to specific autoimmune diseases,
into the peptide binding cleft of HLA-DR and/or DQ proteins

It is believed that the better fit of the peptide determines binding efficiency, and that
this is directly related to T cell negative selection, and T cell
stimulation. An approach that visualized or modeled the peptide sequence
of a protein, with sequential segments fitting into the binding cleft,
should be able to pick out high affinity binding sequences, or possibly
"immunodominant peptides". There is considerable information on the
structure of the DR proteins, the binding cleft, and the conformation of
peptides. There is some work exisitng on the approach that I suggest. A
method that made it relatively easy for researchers to evaluate proteins
involved in autoimmune diseases would have significant impact on medical
research. I can provide more information as needed.

Jimmie Doll (jimmie_doll@brown.edu)
Chemical Physics
Brown University

Numerical path integral methods for "real time" problems.

The math problem underlying the application is performing what amounts to an infinite-dimensional stationary phase integration. the integration variables at issue parameterize physical "paths" that describe the movement of particles through space. since these paths carry phase information, the sum over all such paths gives rise to constructive and destructive interference between the various paths. the "important" paths, the ones that dominate the final results, are thus the for which the phase is "stationary" with respect to local path variation. the result is that the phase interference essentially creates a series of "worm holes" in the three-dimensional space. these "worm holes" evolve in shape/size/number/ as a function of time. to help people understand some of the issues involved, it would be useful to be able to display some of these worm-holes in an intelligent manner.

Elizabeth Brainerd (Elizabeth_Brainerd@brown.edu)
Bio Med Ecology & Evolutionary Biology
Brown University

Extracting 3D motion from CT+moving fluoroscope of animals in motion.

Imaging technologies such as high-resolution CT scanning, MRI, and laser scanning confocal microscopy are opening up vast worlds of cross-sectional and three-dimensional anatomy. In functional morphology and biomechanics, new tools for micrometry, force measurement, 3D flow visualization, 3D motion capture, and mathematical modeling are providing ever more sophisticated understandings of the interactions between morphology and environment. Studies of vertebrate functional morphology, biomechanics, paleontology, and development are poised at the edge of a revolution in our ability to capture and quantify complex morphology and function in 4D (3 spatial dimensions plus time), and to integrate our understandings of function, development, and evolution. In my research program I attempt to take the broadest possible perspective on vertebrate form, function, evolution, and development.

Daniel Acevedo

Last modified: September 6th, 2005