|Joseph O’Sullivan is the Samuel C. Sachs Professor of Electrical Engineering; he also holds a biomedical engineering professorship in the engineering school and a radiology associate professorship in the medical school.
Recognizing a Theory and Its Application
Engineering Professor Jody O’Sullivan builds on information theory for important uses in both the medical and military arenas—from detecting breast cancer via imaging, to identifying camouflaged vehicles from satellites.
On the fourth floor of a building hidden on the north side of campus, across the hall from the secretive-sounding Center for Security Technologies, sits the paper-strewn office of a hunter whose targets are tumors and terrorists. His weapons: information theory, applied math and science, computers, and a wide-ranging mind. His name: Joseph A. “Jody” O’Sullivan, the Samuel C. Sachs Professor of Electrical Engineering and director of the Electronic Systems and Signals Research Laboratory.
The recognition theory and applied imaging science that O’Sullivan develops can be used to locate a tank positioned under a tree via satellite or find cancer lurking in a human brain or breast. The underlying information theory he formulates can help guide an artillery shell or a scalpel. That’s why, in addition to his electrical engineering professorship, he holds a biomedical engineering professorship in the School of Engineering & Applied Science and a radiology associate professorship in the School of Medicine.
“My work is different,” says O’Sullivan, “because I develop theory then applications. Often others don’t see how my theory applies, so I have to be involved. If I don’t see the applications, sometimes no one will.”
His work in the field of recognition, for example, which has dealt in part with pattern recognition, demonstrates his theory-and-application approach.
“In recognition—whether biometrics, recognizing cars or individual faces, or doing brain scans—it all comes from common theory,” says O’Sullivan. “And the various applications send us back to the theory to improve on it.”
Much of O’Sullivan’s foundational work lies in the areas of information theory and information-theoretic imaging.
“Information theory underlies the design and analysis of systems that transmit, store, and process information,” O’Sullivan explains. “It provides bounds and answers what is possible.”
But those bounds may soon be expanding, thanks to work he is immersed in.
Until now information theory has been focused largely on one-dimensional information, says O’Sullivan. That is, information moving along a channel, such as a wire carrying data in and out of a computer. Now, however, he’s developing information theory in two-dimensions and three-dimensions. Instead of moving information merely along a wire, his theory suggests how it might be transmitted and stored on a surface or within a volume—which points to a tremendous potential increase in future information management capability.
But what he’s already developed today stretches our reach in many areas, both medical and military, and others as well.
A medical heritage
O’Sullivan’s father and grandfathers—all three physicians—urged him to enter medicine, but his love of math steered him toward electrical engineering. Ironically, some of his most significant contributions come in the medical field, and their impact on human health may outstrip whatever he might have accomplished as a doctor.
Right now O’Sullivan and fellow researchers have three grants pending with the National Institutes of Health: for hyperspectral imaging of the brain’s surface to extract information about brain function and guide surgery (with Thomas A. Woolsey, primary investigator, professor of experimental neurological surgery and of anatomy and neurobiology at Washington University); PET-CT X-ray imaging for breast cancer detection (with Martin Tornai, primary investigator, associate professor of radiology and biomedical engineering at Duke University Medical Center; and Yuan-Chuan Tai, primary investigator, assistant professor of radiology at Washington University); and optical fluorescence imaging to tag molecules and study metabolic processes (with Joseph Culver, primary investigator, assistant professor of radiology at Washington University).
Hyperspectral imaging uses a computer to analyze optical images across a greatly expanded color spectrum and to detect minute functional differences in given areas. “The goal is to recognize activity,” says O’Sullivan, “to know what different parts of the brain are doing, so you know what you want to cut. I don’t work on where to cut but in recognizing activity.”
One of his collaborators, Thomas A. Woolsey, the George H. and Ethel R. Bishop Scholar in Neuroscience, says that O’Sullivan’s knowledge of image recognition in geological settings, such as identifying camouflaged vehicles from satellite images, could help doctors find their targets on biological terrain.
“Biology changes much faster than geology, so we must speed the image processing up,” says Woolsey. “Jody’s good at translating that knowledge and using math to analyze images that tell you important things about the components. I’m optimistic that this approach will work well in lots of settings.”
Woolsey said image recognition might be used to tell if the blood supply to a particular area has diminished in real time, diagnose skin conditions such as melanoma, or even recognize the presence of foreign substances on the skin—such as explosives.
“Jody’s very effective interacting with others, a must in multidisciplinary research,” says [Professor Donald] Snyder, “and with training and developing students vital to our research. He’s very conscientious about bringing new ideas to his students, constantly evolving the classes he teaches ... ”
Increased security work post-9/11
|Professor Jody O’Sullivan (right) works with Dan Keesing, a second-year doctoral student in biomedical engineering, on Keesing's doctoral research in medical imaging.
Ever since the September 11, 2001, terrorist attacks in New York City and Washington, D.C., a significant part of O’Sullivan’s work has aimed at security issues, to recognize not only military targets but also devise ways of recognizing individuals.
As associate director of the Center for Security Technologies, he works closely with the Center’s director, Ronald S. Indeck, the Das Family Distinguished Professor of Electrical Engineering, and the Center’s assistant director, Robert Pless, assistant professor of computer science and engineering, to encourage, organize, and provide opportunities for research on security problems. The Center supports 40 interdisciplinary collaborators who address fundamental scientific and engineering questions in the design of advanced security systems.
Further, he’s conducting his own research, with his collaborators, not only in spectral analysis for biochemical agent detection, as Woolsey suggests, but also in biometrics, or the statistical analysis of biological data and the recognition, for example, of human fingerprints, faces, and irises.
But all those existing technologies can be compromised, says O’Sullivan, with disguises, contact lenses, or micro-fabrication technology.
So O’Sullivan is seeking alternatives. “We’re exploring the use of other biometric signals that are unique to the individual, measurable, persistent, difficult to replicate, and,” says O’Sullivan, “that definitely demonstrate a particular person’s presence.”
His solution? One that might give the phrase “take your pulse” a whole new meaning.
"We’re looking at heart pulses and pulse shape,” says O’Sullivan, trying to find a way to quantify a biological indicator that may be even more difficult to fake than fingerprints.
Creative problem solving
That sort of creative problem solving is indicative of O’Sullivan’s work, according to colleague Donald L. Snyder, senior professor of electrical and systems engineering.
“He tackles problems in fundamental ways, using math and physics to develop new approaches in a systematic manner,” says Snyder. “He’s made a fundamental contribution to the understanding of steganography,” the science of hiding extra information inside pictures and sounds for secure, clandestine communication.
O’Sullivan devised a theory that sets the limits for the amount of data that can be hidden in a system and provides guidelines on how an adversary might disrupt the hidden information. It has implications not only for copyright protection but also for national security.
“I worked with Pierre Moulin of the University of Illinois on embedding information in digital data sets on video,” says O’Sullivan. “We approached it as a game, the embedder versus the attacker, without assuming that we were smarter than our adversary. Our work has changed the way people approach and define the problem.”
Much of O’Sullivan’s work is collaborative, and much of his success can be attributed not only to his math and science skills but also to his people skills, says Snyder.
“Jody’s very effective interacting with others, a must in multidisciplinary research,” says Snyder, “and with training and developing students vital to our research. He’s very conscientious about bringing new ideas to his students, constantly evolving the classes he teaches with exciting new developments.”
Those collaboration and communication skills led to O’Sullivan’s serving as chair of the Faculty Senate during the 2002–04 academic years and as secretary from 1995–98.
Given all his faculty positions, research interests, teaching, and service to the University, it would seem he’d have little time for much else. But his extracurricular interests are evident in his office, with photos of the five sons he is raising with his wife, Chris, and his basketball trophies.
While continuing to play basketball regularly, O’Sullivan has taken up another athletic challenge: marathon running.
“It’s like research,” says O’Sullivan. “You have a long-range goal and a plan. You build on previous results and ultimately successfully finish the job.”
And as to what that next research challenge is for him, time will tell. Says O’Sullivan: “I get interested in something new each year.”
Rick Skwiot is a free-lance writer based in St. Louis.