血液学指标联合细胞因子的脓毒症免疫麻痹 智能诊断模型构建

Construction of Intelligent Diagnostic Model for Sepsis-Induced Immunoparalysis Based on Combined Hematological Indicators and Cytokines

摘要: 目的：联合血液学指标和细胞因子构建智能诊断模型，对脓毒症免疫麻痹进行早期识别与临床评估。方法：回顾性纳入了150例患者资料，包括健康对照组(25例)、普通炎症组(25例)、脓毒症非免疫麻痹组(50例)及脓毒症免疫麻痹组(50例)。收集所有患者的常规血液学指标、部分炎症和细胞因子指标，逐步进行指标筛选，最后把筛选出来的指标联合起来构建智能诊断模型，用受试者工作特征(Receiver Operating Characteristic, ROC)曲线评估各模型的诊断效能。结果：对患者数据进行筛选，最后选出核心预测指标为：白细胞计数(WBC)、C反应蛋白(CRP)、白细胞介素-1 (IL-1)、白细胞介素-4 (IL-4)。在以上指标的基础上构建智能诊断模型并进行十折交叉验证，结果显示决策树(Decision Tree)模型表现最佳，该模型在训练集与验证集间性能差距小，具有良好的泛化能力，提示决策树模型对脓毒症免疫麻痹具有优异的区分效能。结论：WBC、CRP、IL-1、IL-4是脓毒症免疫麻痹的高效识别标志物，决策树联合模型对其具有判别价值，有助于临床早期识别免疫麻痹状态与及时干预。

Abstract: Objective: To construct an intelligent diagnostic model by combining hematological indicators and cytokines for early identification and clinical evaluation of sepsis-induced immunoparalysis. Methods: A retrospective study was conducted including data from 150 patients, comprising a healthy control group (25 cases), a general inflammation group (25 cases), a sepsis non-immunoparalysis group (50 cases), and a sepsis immunoparalysis group (50 cases). Routine hematological indicators, along with selected inflammation and cytokine markers, were collected from all patients. Stepwise screening of indicators was performed, and the selected indicators were combined to build an intelligent diagnostic model. The diagnostic performance of each model was assessed using the receiver operating characteristic (ROC) curve. Results: Patient data were screened, and the core predictive indicators were ultimately selected as white blood cell count (WBC), C-reactive protein (CRP), interleukin-1 (IL-1), and interleukin-4 (IL-4). Based on these indicators, an intelligent diagnostic model was constructed and subjected to ten-fold cross-validation. The results showed that the Decision Tree model performed best, with a small performance gap between the training and validation sets, indicating good generalization ability. This suggests that the Decision Tree model has excellent discriminative power for sepsis-induced immunoparalysis. Conclusions: WBC, CRP, IL-1, and IL-4 are efficient markers for identifying immune paralysis in sepsis. The combined decision tree model shows good discriminative value, and this could help clinicians recognize immune paralysis early and intervene in time.