To explore the predictive value of CT-enhancement-based spot sign and column line graph models in early hematoma enlargement of cerebral hemorrhage

Wu Fa; Yang Yulin; Wu Tingting; Jiang Rui; Ye Liting; Wang Peng; Huang Ju; Li Jianhao; Du Feizhou

doi:10.3969/j.issn.1002-1949.2024.09.015

Abstract：Objective To deeply investigate the efficacy of spot sign and column line graph(nomogram)model in predicting early hematoma enlargement in the patients with spontaneous cerebral hemorrhage(sICH).Methods This study covered all 210 sICH patients who met the criteria from January 2018 to December 2023 in the Western Theater Command General Hospital of the People's Liberation Army.Based on whether the hematoma was enlarged more than 33％or had a volume of more than 6 mL,the patients were divided into hematoma enlargement group and hematoma non-enlargement group.By comparing and analyzing the clinical characteristics and CT imaging manifestations of patients in the two groups,the risk factors affecting hematoma enlargement were screened by using multivariate Logistic regression analysis.With the help of R language rms package,nomogram model for predicting hematoma enlargement in sICH patients was constructed.The area under curve(AUC)was used to evaluate the differentiation of the model,the calibration curve was used to evaluate the calibration of the model,the decision curve analysis(DC A)was used to evaluate the clinical validity of the model,and the predictive efficacy of the model was comprehensively evaluated by the indexes of accuracy,sensitivity,specificity and Jordon's index.Results There were 66 patients with hematoma enlargement and 144 patients with hematoma non-enlargement in this study.It was found that spot sign,the time of the first computed tomography(CT)examination,hypointensity sign,history of hypertension,homogeneity,hematoma heterogeneity score and fluid level were independent influencing factors in predicting early hematoma enlargement.The AUC value of clinical model was 0.680,with 71.43％accuracy,40.91％sensitivity,88.89％specificity and 0.2980 Jordon's index;the AUC value of spot sign prediction model was 0.838,with 86.19％accuracy,77.27％sensitivity,90.28％specificity and 0.6755 Jordon's index;and the AUC value of nomogram prediction model was 0.910,with 87.14％accuracy,86.96％sensitivity,81.55％specificity and 0.6851 Jordon's index.There was a significant difference among the clinical model,the spot sign model and the nomogram model by DeLong test for all three models.In addition,the nomogram model had a high degree of goodness-of-fit in predicting the probability of early hematoma enlargement versus the actual probability of occurrence.Conclusions The nomogram model constructed in this study based on the spot sign combined with the time of the first CT examination,hypointensity sign,history of hypertension,homogeneity,hematoma heterogeneity score,fluid level has significant value in predicting early hematoma enlargement in the patients with sICH,which provides a strong support for clinical decision-making.