论文部分内容阅读
目的构建COX比例风险预测模型与人工神经网络预测模型,对脑胶质瘤患者术后生存质量进行评价,为临床医师提供简单、准确的评估方法。方法收集2010年6月至2013年8月山西省肿瘤医院收治的58例脑胶质瘤患者的住院治疗及随访资料的年龄、性别、职业等人口学特征,患者入院时的症状、体征、核磁共振成像(magnetic resonance imaging,MRI)检查、病理诊断分型等,肿瘤切除程度、免疫组化检查及Karnofsky功能状态(Karnofsky performance status,KPS)评分等,筛选有意义因素,建立COX比例风险模型,采用预后指数分层和人工神经网络模型,预测患者术后1年生存质量;并采用ROC分析,对两种方法的预测能力进行评价。结果 COX比例风险模型分析表明,伴有癫痫、术前KPS评分、KI67、病理级别、肿瘤切除程度、血供、肢体活动障碍是影响脑胶质瘤患者术后生存质量的主要影响因素。COX比例风险预测模型的灵敏度为60.0%,特异度为83.3%;人工神经网络预测模型的灵敏度为80.0%,特异度为83.3%。结论人工神经网络模型的预测效果优于COX比例风险模型,人工神经网络可为临床医师评价脑胶质瘤患者术后生存质量提供个体化治疗方法。
Objective To construct a COX proportional hazards prediction model and artificial neural network prediction model to evaluate the postoperative quality of life of patients with glioma and to provide clinicians with a simple and accurate assessment method. Methods The data of 58 patients with glioma who were admitted to Shanxi Tumor Hospital from June 2010 to August 2013 were retrospectively analyzed for the demographic characteristics such as age, sex and occupation, the symptoms and signs of patients admitted to hospital, Karnofsky performance status (Karnofsky Karnofsky Karnofsky Karnofsky score, Karnofsky performance status, Karnofsky performance status, KPS) were screened by magnetic resonance imaging (MRI), pathological classification, tumor resection, screening of significant factors, establishment of COX proportional hazards model, The prognostic index stratification and artificial neural network model were used to predict the quality of life of patients one year after operation. The ROC analysis was used to evaluate the predictive ability of the two methods. Results The COX proportional hazards model analysis showed that epilepsy, preoperative KPS score, KI67, pathological grade, degree of tumor resection, blood supply and limb movement disorders were the main influencing factors of postoperative quality of life in patients with glioma. The sensitivity of the COX proportional hazards prediction model was 60.0% and the specificity was 83.3%. The artificial neural network prediction model had a sensitivity of 80.0% and a specificity of 83.3%. Conclusion The artificial neural network model is better than COX proportional hazards model in predicting the quality of life of glioma patients.