Knowledge-based computer-aided detection of masses on digitized mammograms: A preliminary assessment

The purpose of this work was to develop and evaluate a computer-aided detec tion (CAD) scheme for the improvement of mass identification on digitized m ammograms using a knowledge-based approach. Three hundred pathologically ve rified masses and 300 negative, but suspicious, regions, as initially ident ified by a rule-based CAD scheme, were randomly selected from a large clini cal database for development purposes. In addition, 500 different positive and 500 negative regions were used to test the scheme. This suspicious regi on pruning scheme includes a learning process to establish a knowledge base that is then used to determine whether a previously identified suspicious region is likely to depict a true mass. This is accomplished by quantitativ ely characterizing the set of known masses, measuring ''similarity'' betwee n a suspicious region and a ''known'' mass, then deriving a composite ''lik elihood'' measure based on all ''known'' masses to determine the state of t he suspicious region. To assess the performance of this method, receiver-op erating characteristic (ROC) analyses were employed. Using a leave-one-out validation method with the development set of 600 regions, the knowledge-ba sed CAD scheme achieved an area under the ROC curve of 0.83. Fifty-one perc ent of the previously identified false-positive regions were eliminated, wh ile maintaining 90% sensitivity. During testing of the 1000 independent reg ions, an area under the ROC curve as high as 0.80 was achieved. Knowledge-b ased approaches can yield a significant reduction in false-positive detecti ons while maintaining reasonable sensitivity. This approach has the potenti al of improving the performance of other rule-based CAD schemes. (C) 2001 A merican Association of Physicists in Medicine.