|
Tools such as those to be developed at Columbia are expected to help both patients and physicians find reliable health advice that may be a few keystrokes away. The easy accessibility of medical information on-line already is changing the nature of the doctor-patient relationship. (See, for example, "Can the Internet Cure the Common Cold?" in The New York Times, July 9, 1998, p. G1.) Doctors report they no longer need to spend as much time educating patients about their condition, and people afflicted with rare or unusual diseases have been able to locate information on novel therapies unknown to their physicians. But some medical information on the Internet is not peer-reviewed, or is actually wrong. ("On-Line Medical Information Does Poorly on Accuracy Exam," The New York Times, Aug. 5, 1999, p. G3.)
The Columbia project, Personalized Retrieval and Summarization of Image, Video and Language Resources, or PERSIVAL, will create software that will use information from patient records as input for an Internet search. It will extract relevant facts from the records, then seek information in several media that amplifies on the patient's condition. The software will summarize the literature it finds so users can avoid reading many documents. PERSIVAL will use the source of the information - medical association, medical journal, popular magazine, self-help group - to judge its reliability. The software will incorporate language generation techniques that will select everyday English terms, rather than arcane medicalese.
Researchers are assembling a collection of thousands of medical videos - educational materials intended for patients, instructional materials for doctors and diagnostic materials, such as sonograms or echocardiograms - to test PERSIVAL's video search and summary capabilities. The collection, which is to reside at Columbia-Presbyterian Medical Center, will allow patients to match videos of their diagnostic tests with those of normal and abnormal cases, comparing diagnoses.
|
