|
Cover Story |
|
|
Hey, Coach, What's Next?
Computerized coach may observe and | |
BOTH NOVICE AND EXPERIENCED USERS of home-based medical devices may need some coaching to ensure that correct procedures are followed and accurate readings result. That coaching may one day come from a user's home computer.
|
Researchers in Georgia Tech's GVU Center are developing a computerized coach that would observe and instruct users of blood glucose monitors and later be customized for use with other medical devices. The coaching program will use computer vision technology under development by Associate Professor of Computing Irfan Essa and his students. It will not require instrumentation of the user or environment, though a digital video camera would be connected to the user's computer.
"Our ultimate intention is for this system to work for any kind of sequential action," Essa says. "Blood glucose monitors are just one such example. We are interested in any task that employs many different actions and interactions. The end product will be analyzing someone doing a complicated task, guiding them to complete it correctly."
Researcher Casey Fiesler watches study participant Darold Modine use a blood glucose monitor as she simulates the guidance that a computerized coach might give.
(300-dpi JPEG version - 817k)
Researchers have established a methodology to track the steps a person takes using a blood glucose monitor. The system recognizes when the patient has completed, added, substituted and/or omitted a task, such as opening a bottle and inserting a test strip into the device.
One of the biggest hurdles researchers still face is development of artificial intelligence capabilities. Though much of the computer vision technology is now available, the ultimate goal is still dangling like a carrot before Essa and his students Mohammed Hamid and Yan Huang, Essa says.
Meanwhile, School of Psychology Professors Wendy Rogers and Dan Fisk and their students Anne McLaughlin and Casey Fiesler have conducted detailed analyses of what steps should be taken in using the blood glucose monitor, what is an acceptable deviation from the procedure and what constitutes a serious mistake that needs to be corrected. They are also determining how to inform people when they have made a mistake.
In experiments, the psychologists are simulating computer vision with humans who watch patients to see if they can take advantage of feedback they receive. Studies indicate patients do use this feedback to correct their mistakes, and when observed two days later, they are still following the proper procedures. Now, researchers are collecting data from a control group that gets no feedback.
"Blood glucose monitors are difficult to use at first," Rogers notes. "If you don't get feedback, you will probably continue to use it incorrectly or lapse back into a wrong method. The patient might even leave out some steps if they're used to doing it over and over again."
This research is funded by the National Science Foundation, National Institutes of Health and Georgia Tech's
photo by Gary Meek
For more information, contact Irfan Essa, 404-894-6856 or
irfan@cc.gatech.edu;
or Wendy Rogers, 404-894-6775 or
wendy.rogers@psych.gatech.edu.
Contents
Research Horizons
GT Research News
GTRI
Georgia Tech
Send questions and comments regarding these pages to
webadmin@edi.gatech.edu
Last updated: Dec.11, 2003
