Breast cancer risk estimation with artificial neural networks revisited: discrimination and calibration.
Breast cancer risk estimation with artificial neural networks revisited: discrimination and calibration.
复制标题
DOI：
10.1002/cncr.25081
复制
发表时间：
2010-07-15
期刊：
影响因子：
6.2
通讯作者：
Burnside, Elizabeth S.
中科院分区：
文献类型：
--
作者：
Ayer, Turgay;Alagoz, Oguzhan;Chhatwal, Jagpreet;Shavlik, Jude W.;Kahn, Charles E., Jr.;Burnside, Elizabeth S.
关键词：
区分乳腺良恶性病变和准确预测个体患者患乳腺癌的风险是成功的临床决策的关键。在过去，几个人工神经网络（ANN）模型已被开发用于乳腺癌的风险预测。所有这些研究都报告了辨别性能，但没有一项评估了校准，这是准确预测风险的同等重要的措施。在这项研究中，我们评估了人工神经网络（ANN）是否可以区分良性和恶性疾病，并准确预测个体患者乳腺癌的概率。我们的数据集包括62，219个连续收集的乳腺X线摄影结果，与威斯康星州国家癌症报告系统相匹配。我们建立了一个具有1000个隐藏层节点的三层前馈ANN。我们使用十重交叉验证来训练和测试我们的ANN，以预测乳腺癌的风险。我们使用受试者工作特征曲线（AUC）下的面积，灵敏度和特异性来评估放射科医生和我们的ANN的区分性能。我们使用Hosmer-Lemeshow（H-L）拟合优度检验评估了我们的ANN的风险预测（即校准）的准确性。我们的人工神经网络表现出上级的歧视，AUC = 0.965，相比放射科医生，AUC = 0.939（P < 0.001）。我们的ANN也得到了很好的校准，如H-L拟合优度P值为0.13所示。我们的人工神经网络可以有效地区分恶性异常和良性的，准确地预测乳腺癌的风险为个别异常。
Discriminating malignant breast lesions from benign ones and accurately predicting the risk of breast cancer for individual patients are critical in successful clinical decision-making. In the past, several artificial neural network (ANN) models have been developed for breast cancer risk prediction. All of these studies reported discrimination performance, but none has assessed calibration, which is an equivalently important measure for accurate risk prediction. In this study, we have evaluated whether an artificial neural network (ANN) trained on a large prospectively-collected dataset of consecutive mammography findings can discriminate between benign and malignant disease and accurately predict the probability of breast cancer for individual patients. Our dataset consisted of 62,219 consecutively collected mammography findings matched with Wisconsin State Cancer Reporting System. We built a three-layer feedforward ANN with 1000 hidden layer nodes. We trained and tested our ANN using ten-fold cross validation to predict the risk of breast cancer. We used area the under the receiver operating characteristic curve (AUC), sensitivity, and specificity to evaluate discriminative performance of the radiologists and our ANN. We assessed the accuracy of risk prediction (i.e. calibration) of our ANN using the Hosmer–Lemeshow (H-L) goodness-of-fit test. Our ANN demonstrated superior discrimination, AUC = 0.965, as compared to the radiologists, AUC = 0.939 (P < 0.001). Our ANN was also well-calibrated as shown by an H-L goodness of fit P-value of 0.13. Our ANN can effectively discriminate malignant abnormalities from benign ones and accurately predict the risk of breast cancer for individual abnormalities.
登录
查看更多内容
影响因子：
158.5
作者：
Fenton, Joshua J.;Taplin, Stephen H.;Elmore, Joann G.
通讯作者：
Elmore, Joann G.
影响因子：
5
作者：
Cupples, TE;Cunningham, JE;Reynolds, JC
通讯作者：
Reynolds, JC
影响因子：
37.8
作者：
Cook, Nancy R.
通讯作者：
Cook, Nancy R.
DOI：
10.1016/j.jacr.2004.09.012
发表时间：
2005-04-01
期刊：
Journal of the American College of Radiology : JACR
影响因子：
--
作者：
Chan, Evelyn C Y
通讯作者：
Chan, Evelyn C Y
影响因子：
10.3
作者：
Barlow, William E.;White, Emily;Kerlikowske, Karla
通讯作者：
Kerlikowske, Karla