基于机器学习算法的胃印戒细胞癌生存预测研究  

A research for predicting the survival of gastric SRCC patients using machine learning algorithms

作　　者：刘豹 王化琨[1] LIU Bao;WANG Huakun(School of Mathematical Sciences,Heilongjiang University,Harbin 150080,China)

机构地区：[1]黑龙江大学数学科学学院,哈尔滨150080

出　　处：《黑龙江大学自然科学学报》2025年第1期5-12,共8页Journal of Natural Science of Heilongjiang University

基　　金：黑龙江省教育厅科学技术研究项目(12531496)。

摘　　要：胃型胃癌中的印戒细胞癌(Signet ring cell carcinoma,SRCC),具有分化程度低、侵袭性高以及预后差等特点。本研究利用大样本的胃印戒细胞癌患者的临床数据,采用机器学习算法建立生存预测模型,从分类和回归的角度,预测胃印戒细胞癌患者的治愈情况和生存周期。分别使用决策树、随机森林、支持向量机算法和反向传播(Backpropagation,BP)神经网络模型建立分类模型,通过对比各模型在测试集上的预测结果发现,BP神经网络模型的预测准确率达到了82%,召回率为80%,泛化能力最好,模型的预测效果最优。同时,使用加速失效时间(Accelerated failuretime,AFT)与COX回归(COX regression,COX)模型,指标一致性指数(Concordance index,C-index)和赤池信息量准则(Akaike information criterion,AIC)值的评估结果显示,COX回归模型的预测能力较好。最后,使用Kaplan-Meier生存分析算法绘制生存曲线,探究患者生存期概率的影响因素。结果显示,年龄、肿瘤淋巴结转移(Tumor node metastasis,TNM)分期以及手术等因素对患者生存期的概率影响较为明显。The machine learning algorithms are used to build the classification and regression models to predict the cure and survival conditions of gastric signet ring cell carcinoma(SRCC)patients.The data used come from the gastric SRCC clinical data.To build the classification models,the decision trees,random forests,support vector machine(SVM)and back propagation(BP)neural network algorithms are used respectively.By comparing the results on the test data from different models,the accuracy of BP neural network amounts to 82%,and the recall is 80%.The experiment results show that BP neural network achieves the best performance and generalization ability.Simultaneously,the accelerated failure time(AFT)and COX regression(COX)models are used on the clinical data.The results show that COX model has higher values on both concordance index(C-index)and akaike information criterion(AIC).The survival curves by Kaplan-Meier algorithm are plotted,and the survival probability of patients is analyzed.The results show that age,TNM period and surgery have the obvious effects on this probability.

关 键 词：胃印戒细胞癌 机器学习 BP神经网络 生存预测 

分 类 号：O212.1[理学—概率论与数理统计]