预测ER及PR双阴性乳腺癌的CT影像学机器学习模型的建立
Establishment of a CT imaging-based machine learning model for predicting breast cancer negative for both estrogen receptor and pro-gesterone receptor
姜文云 1曹志国 2许志 1周理好2
作者信息
- 1. 皖西卫生职业学院附属医院,安徽六安 237000
- 2. 皖西卫生职业学院
- 折叠
摘要
目的 构建预测雌激素受体(ER)及孕激素受体(PR)双阴性乳腺癌的CT影像学机器学习模型.方法 收集2020年1月—2023年5月本院经病理确诊乳腺癌病人223例,按7∶3的比例随机分为训练集(156例)及验证集(67例).收集病人的临床资料,分别采用CRT决策树及BP神经网络构建ER及PR双阴性乳腺癌的预测模型.结果 年龄、组织学类型、组织学分级、增殖细胞核抗原-67增殖指数、淋巴结转移及毛刺征在双阴性乳腺癌组与非双阴性乳腺癌组间差异有统计学意义(x2=4.078～15.177,P＜0.05).在训练集和验证集中,CRT决策树模型预测双阴性乳腺癌的受试者工作特征曲线下面积(AUC)分别为0.758(95％CI=0.670～0.846)和0.796(95％CI=0.672～0.920),BP 神经网络模型的 AUC 分别为 0.787(95％CI=0.701～0.874)和 0.836(95％CI=0.722～0.950).结论 CRT决策树模型及BP神经网络模型对ER及PR双阴性乳腺癌具有一定的预测效能,BP神经网络模型优于CRT决策树模型.
Abstract
Objective To establish a CT imaging-based machine learning model for predicting breast cancer negative for both estrogen receptor(ER)and progesterone receptor(PR).Methods A total of 223 patients with pathologically confirmed breast cancer in our hospital from January 2020 to May 2023 were enrolled and randomly divided into training group with 156 patients and validation group with 67 patients at a ratio of 7∶3.Related clinical data were collected,and the CRT decision tree model and BP neural network were used to establish a predictive model for breast cancer negative for both ER and PR.Results There were significant differences in age,histological type,histological grade,Ki-67,lymph node metastasis,and spiculation sign between the double-negative breast cancer group and the non-double-negative breast cancer group(x2=4.078-15.177,P＜0.05).The CRT de-cision tree model had an area under the receiver operating characteristic(AUC)curve of 0.758(95％CI=0.670-0.846)in the training group and 0.796(95％CI=0.672-0.920)in the validation group,and the BP neural network model had an AUC of 0.787(95％CI=0.701-0.874)and 0.836(95％CI=0.722-0.950),respectively.Conclusion Both the CRT decision tree model and the BP neural network model have a certain prediction efficiency for breast cancer negative for both ER and PR,and the BP neural network model is superior to the CRT decision tree model.
关键词
乳腺肿瘤/受体,雌激素/受体,孕激素/体层摄影术,X线计算机/机器学习/决策树/神经网络,计算机/预测Key words
breast neoplasms/receptors,estrogen/receptors,progestin/tomography,X-ray computed/machine learning/decision trees/neural networks,computer/forecasting引用本文复制引用
基金项目
安徽高校自然科学研究项目(KJ2021A1363)
安徽高校自然科学研究项目(KJ-2021A1364)
出版年
2024
国家科技期刊平台
NSTL
万方数据