An automatic method to discriminate malignant masses from normal tissue indigital mammograms

Specificity levels of automatic mass detection methods in mammography are g enerally rather low, because suspicious looking normal tissue is often hard to discriminate from real malignant masses. In this work a number of featu res were defined that are related to image characteristics that radiologist s use to discriminate real lesions from normal tissue. An artificial neural network was used to map the computed features to a measure of suspiciousne ss for each region that was found suspicious by a mass detection method. Tw o data sets were used to test the method. The first set of 72 malignant cas es (132 films) was a consecutive series taken from the Nijmegen screening p rogramme, 208 normal films were added to improve the estimation of the spec ificity of the method. The second set was part of the new DDSM data set fro m the University of South Florida. A total of 193 cases (772 films) with 37 2 annotated malignancies was used. The measure of suspiciousness that was c omputed using the image characteristics was successful in discriminating tu mours from false positive detections. Approximately 75% of all cancers were detected in at least one view at a specificity level of 0.1 false positive per image.