CLASSIFICATION OF MICROCALCIFICATIONS IN RADIOGRAPHS OF PATHOLOGICAL SPECIMENS FOR THE DIAGNOSIS OF BREAST-CANCER

Rationale and Objectives. Early detection of breast cancer depends on accurate classification of microcalcifications. We have developed a co mputer vision system that has the potential to classify microcalcifica tions objectively and consistently to aid radiologists in diagnosing b reast cancer. Methods. A convolution neural network (CNN) was used to classify benign and malignant microcalcifications in radiographs of pa thologic specimens. Digital images were acquired by digitizing radiogr aphs at a high resolution of 21 mu m x 21 mu m. Results. Eighty region s of interest selected from digitized radiographs of pathologic specim ens were used for training and testing of the neural network system. T he CNN achieved an A(z) value (area under the receiver operating chara cteristic curve) of 0.90 in classifying clusters of microcalcification s associated with benign and malignant processes. Conclusion. Classifi cation of microcalcifications in pathologic specimens for diagnosis of breast cancer was achieved at a high level in our computer vision sys tem, which consists of high-resolution digitization of mammograms and a CNN.