基于LASSO-随机森林算法的急性脑梗死血管内介入治疗术后出血转化预测模型的构建
杨桂平 , 钟广宏 , 杨少明 , 刘微微
中国现代医学杂志 ›› 2026, Vol. 36 ›› Issue (16) : 27 -35.
基于LASSO-随机森林算法的急性脑梗死血管内介入治疗术后出血转化预测模型的构建
Development of a prediction model for hemorrhagic transformation after endovascular treatment of acute cerebral infarction based on LASSO and random forest algorithms
目的 基于LASSO-随机森林算法构建急性脑梗死(ACI)患者血管内介入治疗术后出血转化的预测模型,并验证其性能。 方法 选取2021年1月—2024年12月河源市人民医院收治的400例行血管内介入治疗的ACI患者作为研究对象,采用完全随机抽样,按7∶3的比例将患者随机分为建模组(280例)和验证组(120例)。收集患者临床资料,采用LASSO回归分析从中筛选关键变量进行多因素一般Logistic回归分析,以筛选出独立危险因素,据此采用随机森林算法构建ACI血管内介入治疗术后出血转化预测模型,分别采用受试者工作特征曲线(ROC)和决策曲线分析(DCA)验证模型的预测性能和临床效用。 结果 该研究纳入的ACI患者术后出血转化发生率为17.00%(68/400)。LASSO回归分析结果显示,基于λ1se,该研究从32个变量中共计筛选出关键变量12个,据此进行多因素一般Logistic回归分析,结果显示:房颤[O^R=4.308(95% CI:1.688,10.998)]、术前SBP水平升高[O^R=1.036(95% CI:1.012,1.062)]、PLT水平降低[O^R=0.979(95% CI:0.968,0.991)]、CRP水平升高[O^R=1.140(95% CI:1.034,1.257)]、桥接治疗[O^R=3.060(95% CI:1.262,7.423)]、PTR延长[O^R=1.059(95% CI:1.021,1.098)]、取栓次数增多[O^R=3.030(95% CI:1.607,5.711)]、使用2种抗栓药物[O^R=3.985(95% CI:1.551,10.236)]、抗栓药物采用负荷剂量[O^R=5.149(95% CI:1.845,14.375)]均为ACI血管内介入治疗术后出血转化的独立危险因素(P <0.05)。ROC曲线分析结果显示,建模组和验证组曲线下面积分别为0.884(95% CI:0.832,0.935)和0.876(95% CI:0.813,0.938)。Hosmer-Lemeshow检验结果显示,两组的模型预测值与实际值比较,差异均无统计学意义(P >0.05)。DCA曲线分析结果显示,在高风险阈值0.05~1.00内,采用随机森林模型对患者进行干预,可获得更高的标准化净收益。SHapley Additive exPlanations分析结果显示,抗栓药物剂量对模型预测的贡献最大,且低特征值会增加出血转化风险;个体预测路径显示,该患者基线风险为0.0496,经各因素累加后最终风险为0.0140。 结论 该研究构建的预测模型能较准确地预测ACI患者血管内介入治疗术后出血转化的发生,可为患者术后治疗方案的制订提供参考依据。
Objective To develop a prediction model for hemorrhagic transformation after endovascular treatment of acute cerebral infarction (ACI) based on LASSO and random forest algorithms, and to verify its performance. Methods Patients with ACI who underwent endovascular treatment at Heyuan People's Hospital between January 2021 and December 2024 were enrolled. Using simple random sampling, the patients were randomly assigned in a 7:3 ratio to a training group (n = 280) and a validation group (n = 120). Clinical data of all patients were collected. LASSO regression was used to identify key variables, which were subsequently entered into multivariable logistic regression analysis to identify independent risk factors. A prediction model for postprocedural hemorrhagic transformation after endovascular treatment for ACI was then developed using a random forest algorithm. Receiver operating characteristic (ROC) curve analysis and decision curve analysis (DCA) were used to evaluate the predictive performance and clinical utility of the model, respectively. Results The incidence of postprocedural hemorrhagic transformation among the 400 patients with ACI was 17.00% (68/400). Based on the λ1se criterion, LASSO regression selected 12 key variables from 32 candidate variables. Multivariable logistic regression analysis showed that atrial fibrillation [O^R = 4.308 (95% CI: 1.688, 10.998) ], higher preoperative systolic blood pressure (SBP) [O^R = 1.036 (95% CI: 1.012, 1.062) ], lower platelet (PLT) count [O^R = 0.979 (95% CI: 0.968, 0.991) ], higher C-reactive protein (CRP) level [O^R = 1.140 (95% CI: 1.034, 1.257) ], bridging therapy [O^R = 3.060 (95% CI: 1.262, 7.423) ], prolonged puncture to recanalization (PTR) time [O^R = 1.059 (95% CI: 1.021, 1.098) ], a greater number of thrombectomy device passes [O^R = 3.030 (95% CI: 1.607, 5.711) ], the use of two antithrombotic agents [O^R = 3.985 (95% CI: 1.551, 10.236) ], and the use of loading doses of antithrombotic agents [O^R = 5.149 (95% CI: 1.845, 14.375) ] were independent risk factors for postprocedural hemorrhagic transformation after endovascular treatment for ACI (all P < 0.05). ROC curve analysis showed that the area under the curve (AUC) values were 0.884 (95% CI: 0.832, 0.935) and 0.876 (95% CI: 0.813, 0.938) in the training and validation groups, respectively. The Hosmer-Lemeshow test showed no significant differences between the predicted and observed outcomes in either group (both P > 0.05). DCA showed that the random forest model provided a higher standardized net benefit across threshold probabilities ranging from 0.05 to 1.00. SHapley Additive exPlanations (SHAP) analysis showed that the dose of antithrombotic medication contributed most to the model predictions, and lower feature values were associated with an increased risk of hemorrhagic transformation. Individual prediction analysis showed that the patient's baseline risk was 0.0496 and decreased to 0.0140 after the cumulative contribution of the individual predictors. Conclusion The prediction model developed in this study accurately predicts the occurrence of postprocedural hemorrhagic transformation after endovascular treatment in patients with ACI and may provide a reference for the formulation of postoperative treatment strategies.
| [1] |
韩建峰. 急性脑梗死机械取栓治疗的研究进展[J]. 西安交通大学学报(医学版), 2025, 46(2): 197-202. |
| [2] |
Duan Q, Li W, Zhang Y, et al. Nomogram established on account of Lasso-logistic regression for predicting hemorrhagic transformation in patients with acute ischemic stroke after endovascular thrombectomy[J]. Clin Neurol Neurosurg, 2024, 243: 108389. |
| [3] |
Chen Y, Lv W, Liu X, et al. Development of a prediction model for hemorrhagic transformation after intravenous thrombolysis in patients with acute ischemic stroke: a retrospective analysis[J]. BMC Med Inform Decis Mak, 2025, 25(1): 227. |
| [4] |
Chen H, Lee J S, Michel P, et al. Endovascular stroke thrombectomy for patients with large ischemic core: a review[J]. JAMA Neurol, 2024, 81(10): 1085-1093. |
| [5] |
Sun T, Yu H Y, Zhan C H, et al. Non-contrast CT radiomics-clinical machine learning model for futile recanalization after endovascular treatment in anterior circulation acute ischemic stroke[J]. BMC Med Imaging, 2024, 24(1): 178. |
| [6] |
中华医学会神经病学分会, 中华医学会神经病学分会脑血管病学组. 中国急性缺血性脑卒中诊治指南2018[J]. 中华神经科杂志, 2018, 51(9): 666-682. |
| [7] |
World Medical Association. World medical association declaration of helsinki: ethical principles for medical research involving human participants[J]. JAMA, 2025, 333(1): 71-74. |
| [8] |
中华医学会神经病学分会, 中华医学会神经病学分会脑血管病学组. 中国急性脑梗死后出血转化诊治共识2019[J]. 中华神经科杂志, 2019, 52(4): 252-265. |
| [9] |
Fu C H, Chen C H, Lin C H, et al. Comparison of risk scores in predicting symptomatic intracerebral hemorrhage after endovascular thrombectomy[J]. J Formos Med Assoc, 2022, 121(7): 1257-1265. |
| [10] |
Ayub T, Barwari A, Finsterer J. Frequency, severity, risk factors, and outcome of hemorrhagic transformation in anterior and posterior stroke[J]. J Clin Med, 2024, 13(7): 2010. |
| [11] |
Ciacciarelli A, Tessitore A, Fiume G, et al. Factors associated with hemorrhagic transformation after endovascular treatment despite early recanalization[J]. J Neurol Sci, 2023, 453: 120778. |
| [12] |
Yu G, Kim Y J, Jeon S B, et al. Thromboelastography for prediction of hemorrhagic transformation in patients with acute ischemic stroke[J]. Am J Emerg Med, 2020, 38(9): 1772-1777. |
| [13] |
Ahmed S R, Zeinhom M G, Ebied A A M K, et al. A multi-center study on the predictors of different subtypes of hemorrhagic transformation of brain infarction after thrombolysis in atrial fibrillation patients presented with embolic stroke[J]. Sci Rep, 2025, 15(1): 15655. |
| [14] |
Maida C D, Norrito R L, Rizzica S, et al. Molecular Pathogenesis of Ischemic and Hemorrhagic Strokes: Background and Therapeutic Approaches[J]. Int J Mol Sci, 2024, 25(12): 6297. |
| [15] |
Tirandi A, Sgura C, Carbone F, et al. Inflammatory biomarkers of ischemic stroke[J]. Intern Emerg Med, 2023, 18(3): 723-732. |
| [16] |
Zang Z, Che S, Liu H, et al. The predictive value of brain CT perfusion imaging and serum CRP for hemorrhagic transformation and clinical outcomes in acute ischemic stroke patients after intravenous thrombolysis[J]. Clin Neurol Neurosurg, 2026, 264: 109347. |
| [17] |
Tian B, Tian X, Shi Z, et al. Clinical and imaging indicators of hemorrhagic transformation in acute ischemic stroke after endovascular thrombectomy[J]. Stroke, 2022, 53(5): 1674-1681. |
| [18] |
Dong S, Yu C, Wu Q, et al. Predictors of symptomatic intracranial hemorrhage after endovascular thrombectomy in acute ischemic stroke: a systematic review and meta-analysis[J]. Cerebrovasc Dis, 2023, 52(4): 363-375. |
| [19] |
Jang K M, Choi H H, Nam T K, et al. Clinical outcomes of first-pass effect after mechanical thrombectomy for acute ischemic stroke: a systematic review and meta-analysis[J]. Clin Neurol Neurosurg, 2021, 211: 107030. |
| [20] |
Pulcine E, Seed M, BrandãO L R, et al. Hemorrhagic transformation and stroke recurrence in children with cardiac disease receiving antithrombotic therapy for secondary stroke prevention[J]. J Thromb Haemost, 2021, 19(10): 2428-2439. |
| [21] |
刘仕庆, 赵凤仙, 周芯羽, 阿司匹林对急性大动脉主干闭塞性脑梗死后癫痫发作的影响[J]. 中国现代医学杂志, 2026, 36(5): 7-13. |
| [22] |
Schnabel R B, Haeusler K G, Healey J S, et al. Searching for atrial fibrillation poststroke: a white paper of the AF-SCREEN international collaboration[J]. Circulation, 2019, 140(22): 1834-1850. |
| [23] |
Thomas S E, Plumber N, Venkatapathappa P, et al. A review of risk factors and predictors for hemorrhagic transformation in patients with acute ischemic stroke[J]. Int J Vasc Med, 2021, 2021: 4244267. |
| [24] |
Chen R, Jiang G, Liu Y, et al. Predictive effects of S100β and CRP levels on hemorrhagic transformation in patients with AIS after intravenous thrombolysis: a concise review based on our center experience[J]. Medicine (Baltimore), 2023, 102(38): e35149. |
| [25] |
Wang S, Jiang J, Gu X, et al. Radiomics signature and deep learning signature of intrathrombus and perithrombus for prediction of malignant cerebral edema after acute ischemic stroke: a multicenter CT study[J]. Front Neurol, 2025, 16: 1650970. |
广东省医学科学技术研究基金项目(B2025210)
河源市科技计划项目(河科社发2025198)
/
| 〈 |
|
〉 |