OncoLink en espanõl
Share
Cancer Resources > Cancer News > Cancer News from Reuters > Reuters Cancer News > 2004 > April
Computer system spots breast lesion misclassification
Last Updated: 2004-04-09 12:00:06 -0400 (Reuters Health)
NEW YORK (Reuters Health) - A computer-aided classification (CAC) system has accurately gauged that many breast lesions deemed "probably benign" by conventional interpretation were in fact malignant, US and Israeli researchers report in the March issue of Radiology.
Dr. Shalom S. Buchbinder of Staten Island University Hospital, New York, and colleagues evaluated the system in a study of 106 cases of lesions.
Although 42 were malignant, blinded retrospective evaluation by at least 2 of 4 radiologists had deemed all these lesions to be of the "probably benign" category 3 of the Breast Imaging Reporting and Data System (BI-RADS).
After training the CAC system on features of digitized mammograms of 646 pathologically proven cases, half of which were malignant, the researchers used it to analyze these 106 cases.
The system incorrectly upgraded the classification of 20 of the 64 benign lesions, giving it a specificity of 69%.
However, it correctly changed the classification of 38 of the 42 malignant lesions. It upgraded 9 to BI-RADS category 4 and 29 to BI-RADS category 5. This yielded a sensitivity of 90%.
Thus, the researchers conclude that assessment "based on CAC methods may help mammographers to appropriately limit their use of the 'probably benign' category."
Radiology 2004;230:820-823.
Copyright © 2009 Reuters Limited. All rights reserved.
Republication or redistribution of Reuters content, including by framing or similar means, is expressly prohibited without the prior written consent of Reuters. Reuters shall not be liable for any errors or delays in the content, or for any actions taken in reliance thereon. Reuters and the Reuters sphere logo are registered trademarks and trademarks of the Reuters group of companies around the world.