Development of early predictive models for malignant cerebral edema based on net water uptake and enhancement ratio from CT

HUANG Lijun; LIAO Anyu; SHEN Xi; CAO Zehong; SHI Feng; ZHU Wusheng; CHENG Xiaoqing; LU Guangming

doi:10.19300/j.2024.L21324

Abstract：Objective Constructing early predictive models for malignant cerebral edema(MCE)following acute anterior circulation large-vessel occlusion based on quantitative imaging metrics,including net water uptake(NWU)and enhancement ratio,derived from CT and CTA source images,and evaluate the model predicting efficiency.Methods A retrospective analysis was conducted on 250 patients with acute ischemic stroke patients due to unilateral anterior circulation large vessel occlusion.According to whether the patients were MCE positive,they were divided into MCE group(73 cases)and non-MCE group(177 cases).NWU,enhancement ratio,and Alberta stroke program early CT score(ASPECTS)of the ischemic area of brain tissue were measured by automated ASPECTS software on CT and CTA source images,and the collateral score were assessed based on Tan scores.Least absolute shrinkage and selection operator(LASSO)and Logistic regression were used to screen the variables and construct two predictive models of MCE based on CT and CTA source images respectively.Internal validation was performed using a 5-fold cross-validation approach.Model performance was evaluated using area under the ROC curves(AUC),specificity,and sensitivity.The Delong test was employed to compare predictive performance between models,calibration curves were generated to evaluate model accuracy,and decision curve analysis(DCA)was utilized to assess clinical utility.Results In comparison to the non-MCE group,the MCE group exhibited higher levels of blood glucose and D-dimer,as well as higher NIHSS scores,NWU,and enhancement ratios.Additionally,the MCE group had lower CT-ASPECTS,CTA-ASPECTS,and collateral scores(P<0.05).LASSO and Logistic regression analysis identified NIHSS score,CT-ASPECTS,CTA-ASPECTS,collateral score,NWU,and enhancement ratio as predictive factors for MCE(P<0.05).Model 1 was constructed based on non-enhanced CT(CT-ASPECTS + NWU)and clinical risk factors(NIHSS),and model 2 was constructed based on single-phase CTA(CTA-ASPECTS + enhancement ratio + collateral score)and clinical risk factors(NIHSS).The AUC of model 1 was 0.800(95%CI:0.743-0.856),the sensitivity was 78.1%and the specificity was 70.6%.The AUC of model 2 was 0.839(95%CI:0.787-0.890),the sensitivity was 80.8%and the specificity was 76.3%.The Delong test indicated no statistically significant difference between the models(P>0.05),while the calibration curve demonstrated that model 2 exhibited a superior goodness of fit,and the decision curve indicated that model 2 yielded a higher clinical net benefit.Conclusion The early prediction models of MCE,established by the automated ASPECTS software calculating NWU on CT and enhancement ratio on CTA source images,demonstrate good prediction ability.A prediction model based on CTA can achieve higher net clinical benefits,and is helpful to promptly screen and identify MCE patients early.