基于MRI的影像组学特征与新生儿缺氧缺血性脑病临床分度间的相关性分析OA北大核心

Background Early identification of children who may develop moderate to severe neonatal hypoxic-ischemic en-cephalopathy(HIE)is a current research focus.The previous reliance on the evaluation of the clinical status of children and other clinical markers has limited effect.Objective To establish a radiomics model by radiomics methods to explore the correlation be-tween the radiomics characteristics of the basal ganglia and thalamus of HIE patients and the clinical grading of HIE.Design Co-hort study.Methods This study retrospectively analyzed 151 HIE patients diagnosed between January 2013 and December 2021,and the grading of HIE was divided clinically.On the T1WI and T2WI sequences of MRI,the region of interest(ROI)was outlined along the maximum area of the basal ganglia and thalamus.Three radiomics labels(based on T1WI,T2WI and the combined mo-dality of T1WI+T2WI,respectively)were constructed through feature screening and dimensionality reduction,and the Rad-score was calculated by linear fitting.Based on the Rad-score of the radiomics model and clinical risk factors,a nomogram was estab-lished to predict mild,moderate and severe HIE.Main outcome measures AUC of the nomogram in the training set and the test set.The calibration curve was used to evaluate the fit between the nomogram prediction of HIE clinical grading and the actual situation.Results Compared with the performance of the radiomics model based on the separate T1WI and T2WI sequences,the radiomics model combining T1WI and T2WI was better.The nomogram constructed based on Rad-score and clinical independent risk factors had an AUC of 0.924(95%CI:0.871-0.997)and 0.849(95%CI:0.704-0.994)in the training group and the test group,respectively.The calibration curve showed that the nomogram had a high fit between the prediction of HIE clinical grading and the actual situation.Conclusion The radiomics features of the MRI basal ganglia and thalamus based on the combination of T1WI and T2WI are related to the clinical grading of HIE.The constructed imaging omics model and nomogram can be used for in-dividualized prediction of mild,moderate and severe HIE with high accuracy.