MALIGNANT AND BENIGN CLUSTERED MICROCALCIFICATIONS - AUTOMATED FEATURE ANALYSIS AND CLASSIFICATION

PURPOSE: To develop a method for differentiating malignant from benign clustered microcalcifications in which image features are both extrac ted and analyzed by a computer. MATERIALS AND METHODS: One hundred mam mograms from 53 patients who had undergone biopsy for suspicious clust ered microcalcifications were analyzed by a computer. Eight computer-e xtracted features of clustered microcalcifications were merged by an a rtificial neural network. Human input was limited to initial identific ation of the microcalcifications. RESULTS: Computer analysis allowed i dentification of 100% of the patients with breast cancer and 82% of th e patients with benign conditions. The accuracy of computer analysis w as statistically significantly better than that of five radiologists ( P =.03). CONCLUSION Quantitative features can be extracted and analyze d by a computer to distinguish malignant from benign clustered microca lcifications. This technique may help radiologists reduce the number o f false-positive biopsy findings.