Data_Sheet_1_Quantitative hematoma heterogeneity associated with hematoma growth in patients with early intracerebral hemorrhage.pdf

BackgroundEarly hematoma growth is associated with poor functional outcomes in patients with intracerebral hemorrhage (ICH). We aimed to explore whether quantitative hematoma heterogeneity in non-contrast computed tomography (NCCT) can predict early hematoma growth.
MethodsWe used data from the Risk Stratification and Minimally Invasive Surgery in Acute Intracerebral Hemorrhage (Risa-MIS-ICH) trial. Our study included patients with ICH with a time to baseline NCCT <12 h and a follow-up CT duration <72 h. To get a Hounsfield unit histogram and the coefficient of variation (CV) of Hounsfield units (HUs), the hematoma was segmented by software using the auto-segmentation function. Quantitative hematoma heterogeneity is represented by the CV of hematoma HUs. Multivariate logistic regression was utilized to determine hematoma growth parameters. The discriminant score predictive value was assessed using the area under the ROC curve (AUC). The best cutoff was determined using ROC curves. Hematoma growth was defined as a follow-up CT hematoma volume increase of >6 mL or a hematoma volume increase of 33% compared with the baseline NCCT.
ResultsA total of 158 patients were enrolled in the study, of which 31 (19.6%) had hematoma growth. The multivariate logistic regression analysis revealed that time to initial baseline CT (P = 0.040, odds ratio [OR]: 0.824, 95 % confidence interval [CI]: 0.686–0.991), “heterogeneous” in the density category (P = 0.027, odds ratio [OR]: 5.950, 95 % confidence interval [CI]: 1.228–28.828), and CV of hematoma HUs (P = 0.018, OR: 1.301, 95 % CI: 1.047–1.617) were independent predictors of hematoma growth. By evaluating the receiver operating characteristic curve, the CV of hematoma HUs (AUC = 0.750) has a superior predictive value for hematoma growth than for heterogeneous density (AUC = 0.638). The CV of hematoma HUs had an 18% cutoff, with a specificity of 81.9 % and a sensitivity of 58.1 %.
ConclusionThe CV of hematoma HUs can serve as a quantitative hematoma heterogeneity index that predicts hematoma growth in patients with early ICH independently.

背景 早期血肿扩大与脑出血（intracerebral hemorrhage, ICH）患者的不良功能预后密切相关。本研究旨在探讨非对比计算机断层扫描（non-contrast computed tomography, NCCT）下的定量血肿异质性是否可预测早期血肿扩大。
方法 本研究使用了急性脑出血风险分层与微创手术（Risa-MIS-ICH）临床试验的数据，纳入基线NCCT检查时间<12 h、随访CT检查时间<72 h的ICH患者。为获取亨氏单位（Hounsfield unit, HU）直方图及亨氏单位变异系数（coefficient of variation, CV），研究人员通过软件的自动分割功能对血肿进行分割，定量血肿异质性以血肿HU的变异系数表征。本研究采用多因素logistic回归分析确定血肿扩大相关参数，通过受试者工作特征曲线下面积（area under the ROC curve, AUC）评估判别评分的预测价值，并依托ROC曲线确定最佳截断值。血肿扩大的定义为：随访CT显示血肿体积较基线NCCT增加>6 mL，或血肿体积相对增幅达33%。
结果 本研究共纳入158例患者，其中31例（19.6%）发生血肿扩大。多因素logistic回归分析显示，基线首次CT检查时间（P=0.040，优势比[OR]=0.824，95%置信区间[CI]：0.686–0.991）、密度分类为“异质性”（P=0.027，OR=5.950，95%CI：1.228–28.828）及血肿HU变异系数（P=0.018，OR=1.301，95%CI：1.047–1.617）均为血肿扩大的独立预测因素。经受试者工作特征曲线评估发现，血肿HU变异系数（AUC=0.750）对血肿扩大的预测价值优于密度异质性分类（AUC=0.638）。血肿HU变异系数的最佳截断值为18%，此时特异性为81.9%，敏感性为58.1%。
结论 血肿HU变异系数可作为定量血肿异质性指标，独立预测早期ICH患者的血肿扩大。