中国实用神经疾病杂志2025，Vol.28Issue(5)：547-551,5.DOI:10.12083/SYSJ.231535
基于随机森林算法的超氧化物歧化酶与脑小血管病总负荷的相关性分析
Correlation analysis of superoxide dismutase and total cerebral small vessel disease burden using random forest algorithm
摘要
Abstract
Objective To investigate the correlation between superoxide dismutase(SOD)and the total imaging burden of cerebral small vessel disease(CSVD)using the random forest algorithm.Methods This study included 181 CSVD patients treated in the Department of Neurology,the First Affiliated Hospital of Air Force Medical University.We randomly split the data into a training set and a test set in 2∶1 ratio.The training set was used to construct a model using the random forest algorithm,while the test set was used to evaluate the model through the receiver operating characteristic(ROC)curve.In addition,we used the SHAP algorithm to visualize the model variables and further analyze the importance and correlation of each variable.Results Univariate analysis showed that age(P=0.01),smoking history(P=0.013),drinking history(P=0.033),high-density lipoprotein(P＜0.001),superoxide dismutase(P＜0.001),homocysteine(P=0.005),cystatin C(P=0.004),and neutrophil-to-lymphocyte ratio(NLR,P=0.002)were statistically significant.The random forest model analysis showed that SOD,NLR,and homocysteine(Hcy)had the highest correlation with the total CSVD burden,with SOD showing a significant negative correlation.The AUC value of ROC of the random forest model in the test set was 0.762,with a 95％ confidence interval of 0.628-0.895.The accuracy was 0.672,the Kappa coefficient was 0.346,the sensitivity was 0.720,the specificity was 0.639,the positive predictive value was 0.581,the negative predictive value was 0.767,the Matthews correlation coefficient was 0.353,and the balanced accuracy was 0.679.Conclusion SOD is significantly negatively correlated with the total CSVD burden and is a protective factor,while NLR and Hcy are significantly negatively correlated and are important risk factors.关键词
脑小血管病/影像学总负荷/超氧化物歧化酶/同型半胱氨酸/随机森林Key words
Cerebral small vessel disease/Total imaging burden/Superoxide dismutase/Homocysteine/Random forest algorithm分类
医药卫生引用本文复制引用
王阳,刘之荣,矫树生,智建霞,欧阳飞,李扬..基于随机森林算法的超氧化物歧化酶与脑小血管病总负荷的相关性分析[J].中国实用神经疾病杂志,2025,28(5):547-551,5.基金项目
国家重点研发计划项目(编号:2017YFC0907703) （编号:2017YFC0907703）
