Prediction Of Gross Motor Function In Children With Spastic Cerebral Palsy Based On MRI Nomogram

Objective: The gross motor dysfunction seriously affects their quality of life and social participation in children with spastic cerebral palsy（SCP）.Therefore,this study explores the responsible lesion associated with gross motor dysfunction and a nomogram was established to assess the degree of gross motor function impairment,so as to improve clinical convenience.Method: SCP children diagnosed by the Affiliated Hospital of Zunyi Medical University were collected from May 2019 to October 2020 and evaluated by Gross Motor Function Classification System（GMFCS）,T₂-weighted fluid-attenuated inversion recovery imaging（T₂-FLAIR）and diffusion tensor imaging（DTI）.（1）GMFCS: Children with SCP were divided into two groups according to GMFCS classification: mild and severe gross motor function impairment,of which GMFCS grades I and II were regarded as mild gross motor function impairment,and GMFCS grades III V were regarded as severe gross motor function impairment.（2）To establish a T₂-FLAIR-based MRI score table to evaluate brain injury in children with SCP.（3）Fractional anisotropy（FA）was obtained by DTI data processing of children with SCP by FSL software.（4）Random forest（RF）was used to screen the characteristics of T₂-FLAIR score items and FA values,and the prediction models of gross motor function were constructed,respectively.Receiver operating characteristic curve（ROC）analysis was used to compare the prediction efficacy between the two models,and the MRI nomogram for predicting gross motor function in children with SCP was further established.The nomogram model was evaluated by calibration diagram and ROC curve.Results: A total of 42 children with SCP was collected,（1）including 17 cases of mild impairment of gross motor function（GMFCS grades I and II）and 25 cases of severe impairment of gross motor function（GMFCS grades III V）,and there were no statistically significant differences in age or gender between the two groups.（2）The score items established based on T₂-FLAIR had good consistency in intragroup and intergroup,with a kappa between 0.64 and 0.96.The area under the curve（AUC）of the gross motor function prediction model based on T₂-FLAIR was 0.94,and the corresponding features were cingulate gyrus,semioval center,corpus callosum,central anterior gyrus,posterior limb of the internal capsule,and lenticular nucleus.（3）The gross motor function prediction model based on FA value its AUC was 0.92,and the corresponding features were the pontine crossing fasciculus,the left superior longitudinal fasciculus,and the middle cerebellar peduncle.Subsequently,the selected features of the two models were fused to construct a gross motor function prediction model,and the results showed that the fused model AUC was 0.93.ROC analysis showed P=0.93 between the gross motor function prediction model based on T₂-FLAIR and the fused gross motor function prediction model.（4）Based on the above,the T₂-FLAIR score was selected to draw a nomogram,consisting of the semioval center,cingulate gyrus,lenticular nucleus,and corpus callosum（OR,0.13,6.69,18.15,and 0.06）in this study.The nomogram AUC was 0.91 and had good calibration.Conclusion: The T₂-FLAIR score is more valuable as a predictor of gross motor function than the FA value.Damage to the centrum semiovale,cingulate gyrus,lentiform nucleus,and corpus callosum can serve as predictors of gross motor function in SCP,which can help determine the responsible lesion.A nomogram composed of T₂-FLAIR-specific site scores can preliminarily predict gross motor function in children with SCP,which in turn helps clinicians to develop treatment plans.