CONNECT: SCIENCE AND VISUALIZATION


Scientific Visualization at NYU: A Color Sampler of Recent Research

by David Frederickson and Estarose Wolfson

[Ed: Links to web pages and/or e-mail addresses which have become inactive since the publication of this article have been enclosed in curly brackets { }. Replacement links have been provided where possible.]

Modern computers are capable of modeling objects—making more or less realistic representations of anything from buildings to molecules to cloud covers—and showing them on the screen or printing them on paper. Scientific visualization can provide images of objects that are real or ideal or theoretical, as large as galaxies, as small as molecules, as abstract as data flows. And color—by now an assumed capability of any computer monitor—provides a dimension that can be seen on the page only when printed in color. Hence this sampler.

The use of color goes beyond making objects look more real or more beautiful, though those of course are valuable objectives. It can also be used to show varying density of solids or gases, or differing atomic structures, or subtle gradations of surface change. In some cases, color can be used to compensate for the relatively low resolution of a computer screen, which is inherently coarser than poor newsprint. In all cases it can make the image more revealing and more intelligible.

A basic problem in planning reconstructive surgery on the face and cranium is deciding upon the best way to cut the bone and reposition the resulting bone fragments. This problem can be solved by simulating surgery before going to the operating room.

A surgical simulator developed by Deljou Khorram Abadi {delijou@mcirps1.med.nyu.edu} under Dr. Court Cutting of the Institute of Reconstructive Plastic Surgery, NYU Medical Center, using the facilities of the ACF's {Scientific Visualization Lab} Replacement URL: http://www.nyu.edu/its/scivis.html, simulates surgery much as it takes place in the operating room, but without the limitations of real surgery. It uses a graphical model of the patient's skull extracted from CT scans; it can cut and reposition each bone fragment interactively, construct postoperative graphics models for evaluation before surgery, and provide surgical plans to be taken to the operating room.

The on-screen simulation can be performed either manually or automatically. For the automatic osteotomies, there is a predefined library of possible incisions. For the automatic optimization, a correspondence is first established between the structural features of the patient's cranium (the ridge above the eye, the edge of the jaw) and a normal one. Based on this, the computer optimizes the movement of each bone fragment to most closely approximate the normative data appropriate for the age, race, and sex of the patient.

In the images to the left, the black mesh represents the normal skull; where the real skull, shown in greenish-gray, is close to normal, the mesh is suppressed. The unmodified cranium is shown first; second, a single osteotomy—only one large piece of facial bone cut and moved—brings slight improvement; at the bottom, a more complex series of eight osteotomies (each shown in a different color) produces a better fit.
The surgical simulator relies on a study that mapped the skull in several hundred thousand triangles—too much data to deal with in a reasonable amount of time. For the purposes of the program, that map was simplified to a few thousand triangles, a much more manageable number. This is a common problem in such research: the need to balance large datasets against constraints of time.

As a postdoctoral scientist in the Department of Computer Science (CIMS), Andre Gueziec {gueziec@watson.ibm.com} (now with IBM) worked with Professor Robert Hummel {hummel@cs.nyu.edu} (CIMS) to develop a Wrapper Algorithm, designed produce an accurate surface modeling of a complex physical object such as a skull or brain. The first stage was to analyze the electronic data from an MR or CT scan as a multitude of minute tetrahedra (rather than the usual cubic voxels, or volumetric pixels), which results in a continuous surface of triangles.
Then the algorithm can be used to simplify the surface: where there is little change from one triangle to its neighbors, they are "grown" into larger triangles; this is done again and again until some of the triangles are quite large (shown to the left in shades of blue); where change is abrupt, the triangles are small (shading through green to yellow and red). The resulting images—shown here both as a mesh of triangles and as the complete surface spanned by the mesh—are both precise and efficient in their representation of the surface geometry.
The other images here have to do with motion—the split-second twisting of supercoiled strands of DNA., the microscopic flows of gases, and the motions of muscle fibers and blood in the heart.
The research of Professor Tamar Schlick (schlick@nyu.edu), of the Department of Chemistry (FAS) and CIMS, deals with the motion of supercoiled DNA in saline solutions, such as those of a normal cell. Variations of salinity can change the behavior of the DNA, which appears to be most flexible at normal cell salt levels. The presence of a solvent in the cell—namely water—affects the forms the supercoiled DNA assumes as it moves. Professor Schlick's molecular simulations seek to depict that behavior over time. Here the colored segments of the supertwisted strand help the viewer to follow the motion; in a series of images representing the same strand in stop-motion over a total of about 1/100,000 second, the loops and segments can be seen to slide past each other. This image was produced using Constantine Kreatsoulas's interface to Per Kraulis's program MolScript, a program that produces pictures of molecules from their Cartesian coordinates; the latter were generated from a simulation done at CIMS by Gomathi Ramachandran {ramachan@Franklin.biomath.nyu.edu}.
Professor Marsha J. Berger (marsha.berger@nyu.edu), also of CIMS, has modeled the flow of a hot dense gas as it leaves a square trench being cut into a medium of low density and temperature. "Ultimately, lasers will be used to dig micron-scale trenches in integrated circuits," she writes. "Before this can be done, it is important to understand the dynamics of the laser-induced flow, so that debris patterns can be categorized or even predicted as a function of energy deposition." In this case, color is used to represent the varying density of the flowing gas.
Professor Charles Peskin (peskin@cims.nyu.edu) of the Department of Mathematics (FAS) and Research Scientist David McQueen of CIMS have developed a computer simulation of a human heart, modeling its motions and the flows of the blood through it. In the paired series of images below, the upper row shows a relaxed heart being filled with blood (oxygen-rich blood from the lungs in red, and oxygen-poor blood from the veins in purple), and the lower row shows a heart contracting to eject the blood. The first image in each row shows the exterior of the heart as a rendered surface; the second is a cutaway view, revealing the blood flow; the third (enlarged), is a wireframe model, indicating the muscular fibrous structure and revealing the flow in greater detail.[ C ]


David Frederickson was the editor of Connect at the time of this article's publications.
{david.frederickson@nyu.edu}
Estarose Wolfson was a Research Scientist with the ACF and oversaw the Scientific Visualization Lab at the time of this article's publication.
estarose.wolfson@nyu.edu

Posted 1 November 1995. Revised 10 July 2007.