Bayesian Network Shows Promise in Mammography

-- Eric Metcalf

Wednesday, June 3, 2009

WEDNESDAY, June 3 (HealthDay News) -- The use of Bayesian networks could allow radiologists to improve their mammographic interpretation, according to research published in the June issue of Radiology.

Elizabeth S. Burnside, M.D., of the University of Wisconsin School of Medicine and Public Health in Madison, and colleagues used reports from 48,744 mammography examinations in 18,270 patients to develop a Bayesian network that was trained to estimate breast cancer risk using demographic risk factors and imaging features cataloged by the radiologist using the Breast Imaging Reporting and Data System lexicon.

The researchers found that the Bayesian network offered significantly better performance compared to the interpreting radiologists in terms of area under the receiver operating characteristic curve (0.960 versus 0.939), as well as sensitivity (90 versus 85.3 percent) and specificity (93.9 versus 88.1 percent).

"The Bayesian network is not intended to replace the radiologist but rather to capitalize on the radiologist's skill in characterizing findings while aiding in the mathematic integration of predictive variables into an accurate risk assessment. In the future, probabilistic computer models like this Bayesian network may substantially aid physicians attempting to diagnose breast cancer in a timely and accurate manner," the authors conclude.

Full Text (subscription or payment may be required)

Specialties Cardiology
Diabetes & Endocrinology
Internal Medicine
Family Practice

Copyright © 2009 ScoutNews, LLC. All rights reserved.


In Celebration of Eric Ott
by Bob Riter
August 17, 2015