基于临床与CT平扫联合模型预测脑出血早期进展的价值

Value of clinical data and non-contrast CT features model in predicting early progression of primary intracerebral hemorrhage

目的:探讨原发性脑出血(pICH)早期血肿扩大的相关高危因素,分别构建临床、影像及两者联合预测模型,评估各模型的诊断价值。方法:回顾性分析208例pICH患者临床、实验室检查、CT平扫资料,依据血肿初始和复查体积的差值,分为稳定组146例,进展组62例。单因素分析筛选2组间有差异的指标,并进一步行多因素logistic回归分析,筛选独立影响因子。分别构建临床、影像和两者联合预测模型,采用ROC曲线和5折交叉验证评估模型的效能及稳定性,DeLong检验评估3种模型AUC的差异,最优模型以列线图展示。利用校准曲线评价最优模型的准确性,决策曲线评价临床获益。结果:2组间性别、脑出血史、低密度脂蛋白胆固醇(LDL-C)、甘油三酯、糖尿病、发病至首次行CT平扫时间(发病时长)、超急性期血肿扩大速度(OTT值)、格拉斯哥昏迷量表(GCS)评分、美国国立卫生研究院卒中量表(NIHSS)评分、初始血肿体积、形态评分、密度分类评分、脑水肿、混合征、黑洞征、岛征、卫星征、旋涡征、低密度征、BAT评分(B为混合征,A为低密度征,T为发病时长)差异均有统计学意义(均P<0.05)。多因素logistic回归分析显示,LDL-C、糖尿病、形态评分、混合征是血肿扩大的独立影响因素。以LDL-C、糖尿病构建临床预测模型;以形态评分、混合征构建影像预测模型;以4个独立影响因素构建联合预测模型。临床、影像和两者联合预测模型的AUC分别为0.671、0.810、0.865,5折交叉验证显示联合预测模型稳定性较好;DeLong检验显示其诊断效能最高(P<0.05);校准曲线显示其一致性较好;决策曲线显示其具有较高的临床实用性。结论:临床联合影像模型在预测脑出血早期进展中价值较高,可为pICH患者的个体化管理提供依据。

Objective:To investigate the related high-risk factors for early hematoma enlargement in primary intracerebral hemorrhage(pICH),and evaluate the diagnostic efficiency of the clinic,imaging and a clinic-imaging models.Methods:Two hundred and eight pICH patients were retrospectively recruited.The clinical and non-contrast CT data were combined for analysis,and all the patients were included in the stability group of 146 cases and the progression group of 62 cases.Single factor analysis was used to screen the indicators with differences between the two groups,and multivariate logistic regression analysis was used to screen the independent influencing factors.The clinic,imaging and clinic-imaging models were constructed separately.ROC curves and 5-fold cross-validation were used to evaluate the effectiveness and stability of the models,and DeLong test was used to evaluate the differences in the AUCs of the three models.The optimal model was shown in the nomogram.The calibration curve was used to evaluate the accuracy of the optimal model,and decision curve analysis was used to evaluate the clinical benefit.Results:There were statistical differences in gender,cerebral hemorrhage history,low-density lipoprotein cholesterol(LDL-C),triglycerides,diabetes mellitus,time from onset to first CT,supraacute onset-to-imaging(OTT)value,Glasgow Coma Scale(GCS)score,National Institutes of Health Stroke Scale(NIHSS)score,initial hematoma volume,morphological score,density score,cephalophyma,blend sign,black hole sign,island sign,satellite sign,vortex sign,low-density sign and BAT score between the two groups(all P<0.05).Multivariate logistic regression analysis showed that LDL-C,diabetes mellitus,morphological score and blend sign were independent influencing factors for hematoma enlargement.And the clinic model was contructed with LDL-C and diabetes mellitus,the imaging model with morphological score and blend sign,the clinic-imaging model with the above 4 influencing factors.The AUC values of the clinic,imaging and clinic-imaging models were 0.671,0.810 and 0.865,respectively.The 5-fold cross-validation showed that the clinic-imaging model was stable,DeLong test showed that had the highest diagnostic efficiency(P<0.05),the calibration curve showed that had a good consistency,and the decision curve showed a high clinical applicability.Conclusions:The clinic-imaging model has a high value in predicting the early progression of cerebral hemorrhage and helps to provide a strong basis for the individual management of patients with pICH.