Clinical-MRI radiomics enables the prediction of preoperative cerebral spinal fluid dissemination in children with medulloblastoma

Abstract Background Medulloblastoma (MB) is the most common pediatric embryonal tumor. Accurate identification of cerebral spinal fluid (CSF) dissemination important in prognosis prediction. Both MRI central nervous system (CNS) and CSF cytology will appear false positive negative. Our objective was to investigate added value preoperative-enhanced T1-weighted image-based radiomic features clinical characteristics predicting preoperative for children with MB. Materials methods This retrospective study included 84 histopathologically confirmed MB between November 2006 2018 (training cohort, n =60; internal validation =24). A set cases December February 2020 were used external ( =40). The normal head spine magnetic resonance images (MRI) no subsequent 1 year diagnosed as non-CSF dissemination. manifested intracranial or intraspinal nodular-enhanced lesions. Clinical collected, conventional examinations evaluated. total 385 extracted from images. Minimum redundancy, maximum correlation, least absolute shrinkage selection operator performed select best performance combined clinical-MRI prediction model developed using multivariable logistic regression. Receiver operating curve analysis (ROC) validate predictive performance. Nomogram decision (DCA) evaluate utility model. Results One nine selected incorporating had training cohort an AUC 0.89. validated cohorts AUCs 0.87 0.73. by a nomogram DCA. Conclusions clinical, MRI, could be applied predict noninvasive biomarker, which aid risk evaluation.