Computer assistance improves lesion classification for experienced mammo readers

Acting as a virtual "second opinion," computer-assisted diagnosis (CAD) could help practitioners differentiate between benign and malignant masses in mammograms with greater consistency, say researchers from the University of Michigan Hospital in Ann Arbor.

Two studies tracked six radiologists as they assessed biopsy-proven masses with a computer classifier that provided the readers with estimates of lesion malignancy. The computer's classification accuracy was better than the average of the radiologists, who had 7-20 years' experience. Thus, when the group read single-view mammograms with the computer aid, all but one radiologist showed improved accuracy in their classification of the masses.

All of the readers were more accurate in classification with two-view mammograms, and further improved with computer assistance. Reading time increased by 30% when the two-view studies were supplemented with CAD, probably because the computer generated separate classification estimates for each view, the authors wrote.

As radiologists grow more comfortable with the CAD, a reduction in the rate of benign mass biopsies is possible, the authors concluded.

To see the full text of this article, visit
www.rsnajnls.org
Page 1 of 240
Next Page