AUTOMATED COMPUTERIZED CLASSIFICATION OF MALIGNANT AND BENIGN MASSES ON DIGITIZED MAMMOGRAMS

Rationale and Objectives. To develop a method for differentiating mali gnant from benign masses in which a computer automatically extracts le sion features and merges them into an estimated likelihood of malignan cy. Materials and Methods. Ninety-five mammograms depicting masses in 65 patients were digitized, Various features related to the margin and density of each mass were extracted automatically from the neighborho ods of the computer-identified mass regions. Selected features were me rged into an estimated likelihood of malignancy by using three differe nt automated classifiers. The performance of the three classifiers in distinguishing between benign and malignant masses was evaluated by re ceiver operating characteristic analysis and compared with the perform ance of an experienced mammographer and that of five less experienced mammographers. Results. Our computer classification scheme yielded an area under the receiver operating characteristic curve (A(z)) value of 0.94, which was similar to that for an experienced mammographer (A(z) = 0.91) and was statistically significantly higher than the average p erformance of the radiologists with less mammographic experience (A(z) = 0.81) (P =.013). With the database used, the computer scheme achiev ed, at 100% sensitivity, a positive predictive value of 83%, which was 12% higher than that for I the performance of the experienced mammogr apher and 21% higher than that for the average performance of the less experienced mammographers (P < .0001). Conclusion. Automated computer ized classification schemes may be useful in helping radiologists dist inguish between benign and malignant masses and thus reducing the numb er of unnecessary biopsies.