论文部分内容阅读
目的运用数据挖掘技术分析与鼻咽癌患者预后有关的各种因素,建立预测鼻咽癌患者5年生存状态的预测模型并评价其效果。方法收集从2005年1月～2007年12月在我院接受诊治的130例鼻咽癌患者的资料。将所有病例分为两组:一组作为训练样本,用于筛选变量及建立预测模型,参与数据挖掘过程,共计104例;一组作为验证样本,用于评价模型效果,不参与数据挖掘过程,共计26例。数据挖掘过程中所采用技术包括单因素分析、logistic回归分析及人工神经网络(ANN)。结果单因素分析显示年龄、T分期、N分期、M分期、92福州分期、KPS评分、颅底骨质破坏、颅神经损伤、咽旁间隙侵犯、确诊到放疗时间、鼻咽疗效、颈部淋巴结疗效共12项指标与鼻咽癌患者的5年生存状态相关(P<0.05)。验证组验证显示,logistic回归模型预测患者5年生存状态的准确率、敏感度和特异度分别为80.8%、81.2%和80%,而ANN模型预测患者5年生存状态的准确率、敏感度和特异度分别为88.5%、87.5%和90%。结论数据挖掘技术可从与鼻咽癌患者预后相关的大量信息中挖掘出有意义的指标,并利用这些指标建立预测模型来判断患者5年后的生存状态。ANN模型的效能优于logistic回归模型。
Objective To analyze various factors related to the prognosis of patients with nasopharyngeal carcinoma by using data mining techniques and establish a prediction model to predict the 5-year survival of patients with nasopharyngeal carcinoma and evaluate the effect. Methods The data of 130 NPC patients who were treated in our hospital from January 2005 to December 2007 were collected. All cases were divided into two groups: one was used as a training sample for screening variables and establishing a prediction model, and involved in the data mining process, a total of 104 cases; a group as a verification sample to evaluate model effects, not involved in data mining process, A total of 26 cases. Techniques used in data mining include univariate analysis, logistic regression analysis and artificial neural network (ANN). Results Univariate analysis showed that age, T stage, N stage, M stage, 92 Fuzhou staging, KPS score, skull base bone destruction, cranial nerve injury, parapharyngeal space invasion, radiotherapy time, nasopharyngeal efficacy, cervical lymph node A total of 12 indicators of efficacy and nasopharyngeal carcinoma patients 5-year survival status (P <0.05). Validation of the validation group showed that the accuracy, sensitivity and specificity of the logistic regression model in prediction of 5-year survival were 80.8%, 81.2% and 80%, respectively. However, ANN model predicts the accuracy and sensitivity of the 5-year survival and The specificity was 88.5%, 87.5% and 90% respectively. Conclusion Data mining can mine meaningful information from a large amount of information related to the prognosis of patients with nasopharyngeal carcinoma and use these indicators to establish a predictive model to determine the patient’s survival status after 5 years. ANN model is better than the logistic regression model.