Process Query System
Gathering information — from network monitors, surveillance cameras, financial records — is easy. Making sense of large quantities of raw data is not.
The Process Query System (PQS), developed by George Cybenko, Dorothy and Walter Gramm Professor of Engineering, employs two tools to solve this challenge: a software framework for categorizing irregularities within a system and algorithms that can provide detailed explanations of those irregularities.
All computer systems are characterized by distinct states, dynamics, and other properties that can be picked up by sensors. Installed into a particular system, PQS quickly reads such data, detects changes or oddities, and infers intent by separating honest errors from potentially deceptive activities.
Perhaps the most useful PQS application to date is in the area of network security. Currently available monitoring tools produce information in quantities that make analysis a formidable task. “PQS closes the gap between gathering a tremendous amount of valuable data and figuring out what the data mean,” says Professor Cybenko.
Other applications for PQS could include scanning credit reports for identity theft or detecting suspicious activity at an international border.
Cybenko and research associate and lecturer Vincent Berk published a paper on PQS in the January 2007 issue of IEEE Computer. The research, a project of Dartmouth’s Institute for Security Technology Studies, is supported in part by funding from the U.S. Department of Homeland Security, Directorate for Science and Technology, and the Department of Defense (DTO, AFRL, and DARPA).
The efficiency of current power electronics equipment is limited by the typical power loss in high frequency power converters. Doctoral candidate Jennifer Pollock and Professor Charles Sullivan have recently patented a new inductor technology that is likely to become the industry standard for converters used in hybrid vehicles, wind energy systems, photovoltaics, fuel cells, and other applications.
The new technology provides the low DC resistance of a foil-wound inductor without the high AC resistance ordinarily found in such inductors. The result is an inductor that is both smaller and more efficient than inductors currently used in power supplies, inverters, and electric motor controllers. The technology is well suited to newer silicon devices such as Insulated Gate Bipolar Transistor (IGBT) modules, which handle currents in the order of hundreds of amperes with blocking voltages of up to 6000 volts while operating at frequencies over 10 kHz.
Tests in Professor Sullivan’s labs measured the potential energy savings of the new foil-winding technology and found that total winding losses of the new inductor were 17–35% lower than those in conventional solid-wire or litz-wire inductors.
The technology has been licensed by West Coast Magnetics in Stockton, Calif. The company forecasts a global market for inductors using this technology that may reach $2.5 billion by 2015 with the greatest opportunities occurring in the hybrid vehicle market and in power generation for wind and solar power equipment. If the technology is extended into lower power applications such as personal computers, desktop electronic equipment and handheld devices, West Coast Magnetics says that the 2015 global market could exceed $5 billion.
Advances in Breast Cancer Detection
By combining two techniques, magnetic resonance imaging (MRI) and near-infrared optics (NIR), researchers led by Professor Keith Paulsen may have devised a new, potentially more accurate method for diagnosing breast cancer. Their pilot study, demonstrating the feasibility of the concept, was published in the April 15 issue of the journal Optics Letters, published by the Optical Society of America.
MRI produces information on the shape and composition of breast tissue, while NIR measures its blood volume and oxygen saturation. Thus the MRI provides information on the form and the NIR on the function. Together the two techniques create high-resolution functional images of tissue, which can then be compared with tissue that is known to be cancerous.
The pilot study involved a 29-year-old woman with a ductal carcinoma, a common breast cancer. Using the information from a contrast MRI procedure — one MRI done before and one after the contrasting agent gadolinium is injected — the research team pinpointed the region for the NIR. Results showed tissue with high hemoglobin level, low oxygen saturation, and high water content — all indicators of cancerous tissue.
A follow-up study will draw on volunteers who have breast abnormalities and have been recommended for biopsy. Using the MRI/NIR technique on the subjects before and after the biopsy, the researchers will be able to compare their results to the biopsy results.
The Thayer researchers, including Paulsen, Professors Brian Pogue and Shudong Jiang, and Adjunct Professsors Hamid Dehghani and John Weaver, are collaborating with Dartmouth Medical School researchers and Dartmouth-Hitchcock Medical Center clinicians.