Cancer Prediction of breast cancer malignancy using an artificial neural network英语教案.pdfVIP

2944 Predictionof Breast Cancer Malignancy Using an Artificial Neural Network Carey E. Floyd,Jr., Ph.D.,*,tJoseph Y. Lo, Ph.D.,*,tA. Joon Y u n ,M.D.,* Daniel C. Sullivan, M.D.,* and Phyllis J. Kornguth, M.D., Ph.D.* Background. An artificial neural network (ANN)was technique for early detection of breast cancer, a signifi- developed to predict breast cancer from mammographic cant fraction of patients referred for biopsy as the result findings. This network was evaluated in a retrospective of mammography findings do not have a malignancy. study. Although specific, biopsy is an invasive, costly, and Methods. For a set of patients who were scheduled emotionally stressful procedure. In an effort to reduce for biopsy, radiologists interpreted the mammograms and the number of benign cases sent to biopsy, an artificial provided data on eight mammographic findings as part intelligence technique was investigated to predict the of the standard mammographic workup. These findings were encoded as features for an ANN. Results of biopsies outcome of biopsy from radiographic findings. An arti- were taken as truth in the diagnosis of malignancy. The ficial neuraf network (ANN) was developed that takes ANN was trained and evaluated using a jackknife sam- radiologic findings as inputs and predicts the outcome pling on a set of 260 patient records. Performance of the of biopsy as an output. network was evaluated in terms of sensitivity and speci- ANNs are comput