Radiomics Based of Deep Medullary Veins on Susceptibility-Weighted Imaging in Infants: Predicting the Severity of Brain Injury of Neonates with Perinatal Asphyxia

Objective: This study aimed to apply radiomics analysis of the change of deep medullary veins (DMV) on susceptibility-weighted imaging (SWI), and to distinguish mild hypoxic-ischemic encephalopathy (HIE) from moderate-to-severe HIE in neonates. Methods: A total of 190 neonates with HIE (24 mild HIE and 166 moderate-to-severe HIE) were included in this study. All of them were born at 37 gestational weeks or later. The DMVs were manually included in the regions of interests (ROI). For the purpose of identifying optimal radiomic features and to construct Rad-scores, 1316 features were extracted. LASSO regression was used to identify the optimal radiomic features. Using the Red-score and the clinical independent factor, a nomogram was constructed. In order to evaluate the performance of the different models, receiver operating characteristic (ROC) curve analysis was applied. Decision curve analysis (DCA) was implemented to evaluate the clinical utility. Results: A total of 15 potential predictors were selected and contributed to Red-score construction. Compared with the radiomics model, the nomogram combined model incorporating Red-score and urea nitrogen did not better distinguish between the mild HIE and moderate-to-severe HIE group. For the training cohort, the AUC of the radiomic model, and the combined nomogram model were 0.84, 0.84. For the validation cohort, the AUC of the radiomic model, and the combined nomogram model were 0.80, 0.79. The addition of clinical characteristics to the nomogram failed to distinguish mild HIE from moderate-to-severe HIE group. Conclusion: We developed a radiomics model and combined nomogram model as an indicator to distinguish mild HIE from moderate-to-severe HIE group.