 | | | | Related Healthscout Videos | |
|
By Vivian Richardson, Ivanhoe Health Correspondent
ORLANDO, Fla. (Ivanhoe Newswire) -- A new way to sift through thousands of mammograms could one day help catch more breast cancers and cut down on the number of false positives -- all in less time than it takes now.
Text Continues Below
Now, computer-assisted detection (CAD) systems compare a new mammogram image to thousands of previously collected images and help radiologists decide whether or not a suspicious area is actually a tumor. The problem? A growing number of images in the database are bogging down the process. Researchers at Duke University in Durham, N.C., wanted to come up with a way to speed it up. What they devised is a bit like an Internet search engine. It classifies what is the most important data to look at and ignores the rest.
"Just like any search engine needs a fast strategy, a smart strategy to sift through the data fast, we needed something like that for our own system," said Georgia D. Tourassi, Ph.D., lead researcher of the study, today at the 48th Annual Meeting of The American Association of Physicists in Medicine. Instead of comparing the new mammogram to every mammogram in the database, she and her colleagues designed a program that compares it to a smaller subset of the database. The images contained in this smaller subset are the most informative.
The researchers tested the new method on a database of 2,318 sections of mammographic images containing 905 masses and 1,413 normal segments. Using both the traditional system -- comparing a test image to every image in the database -- and the new method -- only comparing the test image to a specified subset of 600 images -- researchers report they were able to achieve the same level of detection performance.
Dr. Tourassi said, although this new method is only in the early stages of development, this system, or one like it, could help radiologists be more confident in their diagnosis. "It's like having the 20 to 30 years of experience a radiologist acquires over time sitting there in a knowledge database they can access," she said.
SOURCE: Vivian Richardson at the 48th Annual Meeting of The American Association of Physicists in Medicine in Orlando, Fla., July 30-August 3, 2006
This article was reported by Ivanhoe.com, who offers Medical Alerts by e-mail every day of the week. To subscribe, go to: http://www.ivanhoe.com/newsalert/.
|
.gif)
