By Lissa Harris
With research subjects as large as Mars and as small as a strand of RNA, graduate students had a chance to showcase the results of their work in the Statler Hotel ballroom March 29.
The occasion was the second annual Engineering Graduate Research Symposium sponsored by the College of Engineering and the Engineering Graduate Student Association. On display were poster presentations on 32 research projects. Three of the presenters were chosen to speak at the symposium dinner, each speaker receiving an award of $400. A judging committee also selected the five students with the best posters to receive awards of $200 each. Michael Hayes, director of research, graduate studies and professional education in the engineering college, sees a strong future for the annual symposium. Said Hayes, "We're hoping that this is going to become bigger and better, to be an outstanding tradition within the college." Dinner speaker Jeffrey Lee, a student in electrical and computer engineering, has developed an electrocardiogram (EKG) heart monitor that can send data in real time over a wireless Ethernet connection. Lee sees the use of wireless Ethernet as a means to expand the amount and the resolution of data captured by EKG monitors, which keep track of the electrical impulses generated by a beating heart. With more data at their disposal, he said, cardiologists will be able to develop better predictions about who is at risk for a heart attack.
"Most of the commercial halter devices [the portable EKG monitors most commonly used today] use somewhere between 2 and 5 megabytes of data storage a day," he said. "In order to do sophisticated data analysis, you need higher resolution, a better set of data. Our scheme uses approximately 2 gigabytes a day." The wireless EKG device also measures the heartbeat at a higher frequency than does a standard device -- about 5,000 times per beat, in contrast to the standard device's 300 times per beat. Kiri Wagstaff, another symposium speaker, is interested in computers that learn. Wagstaff, a student in the Department of Computer Science, has developed ways to make "learning" algorithms more intelligent. Clustering algorithms, which automatically sort data into groups based on similarity, can process extremely large amounts of data that would overwhelm a human researcher. But because the algorithms do not have knowledge of the subject matter they analyze, said Wagstaff, they make mistakes that humans would never make.
"The clustering algorithm does not know what it is clustering, and it doesn't know what the features mean. It will perform the same analysis regardless of whether I give it geometric shapes, or apples, or census data on people or data on observations of stars," she said. However, Wagstaff has developed a way to make an algorithm a "temporary expert," giving it preferences for grouping objects in particular ways and, thus, building into the learning process some of the expertise a human researcher would bring to the task of analyzing data. Wagstaff has used intelligent clustering algorithms to tackle very different problems -- decoding English sentences, creating road maps that are accurate enough for an automatic "robo-car" to follow and identifying Martian landscape features in images acquired by the Hubble Space Telescope. Sylvia Kwakye, a graduate student in biological and environmental engineering, spoke about creating micro-devices that can detect the presence of disease-causing organisms. The need for these devices has become particularly apparent, she said, in light of heightened concern about bioterrorism and the need for rapid detection of pathogens such as anthrax. Inside one of Kwakye's devices, tiny "capture" and "reporter" probes, are suspended in fluid over a magnet. The capture probes are attached to magnetic beads, while the reporter probes are attached to microscopic spheres called liposomes that contain marker molecules. If these probes encounter a piece of RNA from the particular organism they are designed to detect, they form a layer on the surface of the magnet, with RNA strands sandwiched between the capture probes and the reporter probes. Lipsomes attached to reporter probes then release their marker molecules, which are electrochemically active and thus generate a measurable electrical current in the device.
The five poster award winners were Junho Bae in mechanical engineering, Phong Du in materials science and engineering, Bin Gao in biological and environmental engineering, Erin Iesulauro in civil and environmental engineering and Reid Orth in biomedical engineering.