August 17, 2009 –
Kalashnikov
A lunchtime audience last week became acquainted with two more SURF-IT research projects as the summer seminar series continued.
Dmitri Kalashnikov, assistant adjunct professor of computer science and researcher on the NSF-funded ResCUE emergency management project, detailed the ways in which identity in multimedia events can be detected from data gathered by the 200 sensors installed in Bren Hall’s “smart spaces.”
The 10 different types of sensors in the sentient space collect a wide range of data, and researchers rely on semantic middleware called SatWare to identify and analyze it. Because data analysis is only as good as the data itself, it’s important to clean and prepare data before working with it. Kalashnikov, who works with computer science professor Sharad Mehrotra, is focusing specifically on what he calls disambiguation – removing the ambiguity that results when object descriptions or references are not unique identifiers.
Researchers modified a framework called Graph Disambiguation Framework (GDF) that was originally developed to parse information quality in database references. The adapted GDF can analyze raw data, including features like hair, clothing color, time and location in order to identify individuals in specific events.
The domain-independent framework relies upon the observation that many real-world datasets are relational in nature and contain not only information about entities but also about the relationships among them. It is based on finding connections in entity-relationship graphs, much like finding connections in social networks, Kalashnikov said. Features and event specifics are analyzed for similarities, and linked by the strength of those connections.
Researchers are also applying the methodology to the problem of Web entity search, a way to group web pages. When the name John Smith is searched, the result would be hundreds of Web pages that mention that name. Using the GDF would allow the user to group the pages that refer to the same person – again, based on the strength of the connections in the graphs.
GDF is a new paradigm, Kalashnikov said. It is domain-independent, self-tuning, scalable to large datasets, robust and has a high separation quality. More information: www.ics.uci.edu~dvk/GDF.
Yunan Chen, informatics assistant professor, is interested in medical informatics, specifically human/computer interactions and the integration of medical records. She is researching the ways in which people use and manage their own health information with the goal of designing software to help them better manage these tasks, which, she believes, will improve overall health.
Chen champions the concept of personal health information management, which she believes leads to better decision-making and optimal healthcare choices.
“Patients are the most under-utilized resource in the healthcare delivery system,” she said. “The goal is to empower consumers and keep them informed about their health issues.” Achieving this requires not only better personal record-keeping but collaboration between patients and physicians. “Patients have to be actively learning about their health issues, so when physicians offer you [treatment] you have some information,” she said.
Chen
In the U.S., Chen says, it is normal for healthcare providers to keep patients’ medical records, but in other countries, patients obtain their records and store them at their home. Chen is interested in learning what they do with this information when they have access to it?
She framed the question using two terms: personal health information management and personal health records.
Personal health information management involves knowing where to find medical information when a patient needs it. Chronic patients may have multiple diseases and visit lots of doctors, she said, and their medical information is often fragmented because it’s stored in different places.
Chen has found that a lot of patients accumulate health information but have difficulty finding it when they need it. In addition, there is so much information available regarding certain medical conditions that patients often have difficulty absorbing it all and integrating it into their lives.
Another aspect of consumer medical informatics is personal health records – an online system that allows individuals to log on and view their medical records online. There are three types of these systems: commercial PHR systems, which include services like Google Health in which patients enter their own data; linked PHR, which provide continuity of medical records systems inside large healthcare providers like Kaiser; and an integrated system, which allows patients to access a lifetime worth of medical records from wherever they are. The integrated system is PHR’s ultimate goal, Chen said, but it is still unrealized.
Chen’s research has found that even when patients have access to their medical records they don’t really read or use it. They use it mainly to support care transitions by taking it with them when they see new doctors. Even this small step, however, makes the patient more active in their own health management.
Ultimately, she would like to see PHR and PHIM systems co-joined. “There is a huge opportunity for people to integrate these two parts of health management together,” she said.
Chen’s current project delves into learning how patients use technologies to manage their health information. How do they gather and use information in healthcare and home settings? How can technology support care transitions from multiple settings and support information sharing? She and her SURF-IT student, James Milewski, are studying patients with Type-2 diabetes; they have completed nine semi-structured interviews conducted in patients’ home environments, and are hoping to have 15 patients in the study by the end of the month. Their next step is to code their data and design a prototype to support health information management activities.
