脑室周围白质损伤患儿发生混合型脑性瘫痪预测模型的建立

Development of A Model to Predict Mixed Cerebral Palsy in Infants with Periventricular White Matter Injury

目的建立脑室周围白质损伤(PWMI)患儿发生混合型脑性瘫痪的预测模型。资料与方法回顾性纳入2015年9月—2022年10月河南中医药大学第一附属医院经MRI诊断为PWMI(6个月~2岁)并随访至2岁后确诊为混合型脑性瘫痪的患儿,将其分为混合型组和痉挛型组。使用多因素Logistic回归分析筛选与PWMI混合型脑性瘫痪相关的MRI征象并建立预测模型,采用五折交叉及重复交叉验证对模型进行内部验证。评估模型的区分度、校准度并进行决策曲线分析。分析独立MRI征象与混合型脑性瘫痪粗大运动功能分级的相关性。结果共纳入135例PWMI脑性瘫痪患儿,痉挛型100例,混合型35例。丘脑腹外侧核(OR=27.500,95%CI 8.293~90.942)、后壳核(OR=13.700,95%CI 4.489~41.549)、海马(OR=7.200,95%CI 1.702~30.813)及尾状核损伤(OR=5.800,95%CI1.973~16.950)与混合型脑性瘫痪相关,并基于以上4个变量构建预测模型。模型的曲线下面积为0.960(95%CI0.934~0.988),五折交叉及重复交叉验证的曲线下面积分别为0.95、0.96;并具有良好的校准度(χ^(2)=3.712,P=0.529)及临床应用性。模型的4个独立MRI征象均与粗大运动功能分级相关(r=0.559、0.581、0.171、0.409,P均<0.05)。结论本研究建立的预测模型可早期准确地预测PWMI混合型脑性瘫痪高危患儿。

Purpose To development a model to predict mixed cerebral palsy(CP)in infants with periventricular white matter injury(PWMI).Materials and Methods This study retrospectively included infants with PWMI on MRI aged 6 to 24 months and were diagnosed as CP after age of 2 years in the First Affiliated Hospital of Henan University of Chinese Medicine,from September 2015 to October 2022.The eligible infants were divided into mixed group and spastic CP group.Multivariable Logistic regression model was used to select the MRI features associated with PWMI and mixed CP,and was internal validated by using the five-fold cross and repeated cross validation.Model performance was evaluated by the discrimination,calibration and decision curve.The correlation between independent MRI features and the gross motor function classification system levels was evaluated.Results A total of 135 infants with PWMIand CP were included in this study,100 with spastic CP,and 35 with mixed CP.Multivariable Logistic regression analysis found that the involvement of ventralateral of thalamus(OR=27.500,95%CI 8.293-90.942),posterior putamen(OR=13.700,95%CI 4.489-41.549),hippocampus(OR=7.200,95%CI 1.702-30.813)and caudate nucleus(OR 5.800,95%CI 1.973-16.950)were associated with PWMI and mixed CP.A prediction model was constructed using the above four MRI features.The model yielded an area under the receiver operating characteristic curve of 0.960(95%CI 0.934-0.988)and 0.95,0.96 in the five-fold cross and repeated cross validation,respectively.The model had good calibration(χ^(2)=3.712,P=0.529)and clinical application.Furthermore,the four MRI variables were associated with gross motor function classification system levels(r=0.559,0.581,0.171,0.409,all P<0.05).Conclusion The model can early and accurately predict infants with PWMI at high risk of mixed CP.