Computerized classification of malignant and benign clustered microcalcifications in mammograms

The purpose of this study was to evaluate the performance of the authors' computerized classification scheme for clustered microcalcifications using two independent databases. The computer scheme estimates the likelihood that a microcalcification cluster is malignant on the basis of eight computer-extracted image features using an artificial neural network. Two biopsy-proven microcalcification databases were used in the performance evaluation, one of which was a quasi-consecutive biopsy series. The classification performance of the computer scheme was compared to the performance of two groups of five radiologists. On both databases, the classification performance of the computer scheme was statistically significantly better than that of the radiologists. This study demonstrates the potential of the computer scheme in clinical applications.