PERSEUS is a new medical search engine


New member
Reprinted from The Targum

For patients who seek medical advice online, Google may soon prove to be outdated.

Am?lie Marian, a University computer science assistant professor, and No?mie Elhadad, a biomedical informatics assistant professor from Columbia University, are developing a new search engine for patient-written content called PERSEUS.

"When someone searches for ?dry mouth,' is that the symptom of an ailment or the side effect of a treatment?" Marian said. "We want to let patients define exactly what it is they're looking for, and then use natural-language processing techniques to get that information."

PERSEUS, funded by the National Science Foundation and a Google Research Award, will allow users to take advantage of filters to provide the accurate information they seek from established online forums, she said.

The tool, which will be completed in two to three years, features a search engine that identifies textual patterns and structures to return relevant results, Elhadad said.

"You can search for others who have had your condition and read about their experiences or search for information about a treatment and what people have to say about it," she said.

Elhadad said PERSEUS' advantage lies in the trust patients have in online communities.

While patients may withhold personal information they consider embarrassing from their physicians, they are more candid in discussing their health problems online under a veil of anonymity, she said.

"Our goal is to help patients find emotional support in online communities and to let them rely on each other for specific information about their symptoms or a treatment," Elhadad said. "It's not meant to replace a medical expert, but it's easy to use and anyone can use it."

Health care professionals will be able use the tool to find common concerns patients have about treatments or symptoms and be better prepared to address them during visits, Marian said.

Scientific researchers can use PERSEUS to obtain patient feedback about prescriptions or medical instruments, she said.

Elhadad said some obstacles the project faces include the reliability of information obtained from online forums and the challenge of analyzing language in that context.

"Obviously, not all information from online forums is going to be correct," Elhadad said. "But we try to search in large online communities where errors are usually rectified by other members within a few hours. It will be the patient who decides what advice is reliable and what isn't."

Online forums often contain misspellings, loose grammar and slang, which poses a difficulty for computerized natural-language processing, she said.

By designing improved algorithms for machine-learning and statistical pattern recognition, Elhadad made the search tool recognize these aberrations.

"Unlike news stories or dedicated texts written by medical experts, people posting on forums have no uniform voice," she said. "The drug Tamoxifen might be called by its brand names Tamofen or Tamone or might even be misspelled. We need to program the tool to pick up on these deviations."

Ethan Jiang, a School of Arts and Sciences sophomore, thinks the tool will be helpful in diagnosing himself when he's sick.

"Whenever I feel sick and don't know what the problem is, I always search online first," he said. "If it doesn't seem serious, I could save myself a trip to the doctor. If this tool can make searches easier, then all the better."

(Note: A copy of this original news story may be obtained from her