Zheng Qiang, Martin-Saavedra Juan Sebastian, Saade-Lemus Sandra, Vossough Arastoo, Zuccoli Giulio, Gonçalves Fabrício Guimarães, Freeman Colbey W, Ouyang Minhui, Singh Varun, Padula Michael A, Demauro Sara B, Flibotte John, Eichenwald Eric C, Detre John A, Sze Raymond Wang, Huang Hao, Hwang Misun
Yantai University, Yantai, China.
Children's Hospital of Philadelphia, Philadelphia, PA, United States.
Front Pediatr. 2020 Sep 25;8:576489. doi: 10.3389/fped.2020.576489. eCollection 2020.
To compare cerebral pulsed arterial spin labeling (PASL) perfusion among controls, hypoxic ischemic encephalopathy (HIE) neonates with normal conventional MRI(HIE/MRI⊕), and HIE neonates with abnormal conventional MRI(HIE/MRI⊖). To create a predictive machine learning model of neurodevelopmental outcomes using cerebral PASL perfusion. A total of 73 full-term neonates were evaluated. The cerebral perfusion values were compared by permutation test to identify brain regions with significant perfusion changes among 18 controls, 40 HIE/MRI⊖ patients, and 15 HIE/MRI⊕ patients. A machine learning model was developed to predict neurodevelopmental outcomes using the averaged perfusion in those identified brain regions. Significantly decreased PASL perfusion in HIE/MRI⊖ group, when compared with controls, were found in the anterior corona radiata, caudate, superior frontal gyrus, precentral gyrus. Both significantly increased and decreased cerebral perfusion changes were detected in HIE/MRI⊕ group, when compared with HIE/MRI⊖ group. There were no significant perfusion differences in the cerebellum, brainstem and deep structures of thalamus, putamen, and globus pallidus among the three groups. The machine learning model demonstrated significant correlation ( < 0.05) in predicting language( = 0.48) and motor( = 0.57) outcomes in HIE/MRI⊖ patients, and predicting language( = 0.76), and motor( = 0.53) outcomes in an additional group combining HIE/MRI⊖ and HIE/MRI⊕. Perfusion MRI can play an essential role in detecting HIE regardless of findings on conventional MRI and predicting language and motor outcomes in HIE survivors. The perfusion changes may also reveal important insights into the reperfusion response and intrinsic autoregulatory mechanisms. Our results suggest that perfusion imaging may be a useful adjunct to conventional MRI in the evaluation of HIE in clinical practice.
比较对照组、常规MRI正常的缺氧缺血性脑病（HIE）新生儿（HIE/MRI⊕）和常规MRI异常的HIE新生儿（HIE/MRI⊖）的脑动脉搏动式自旋标记（PASL）灌注情况。利用脑PASL灌注建立神经发育结局的预测机器学习模型。共评估了73例足月儿。通过置换检验比较脑灌注值，以确定18例对照组、40例HIE/MRI⊖患者和15例HIE/MRI⊕患者中灌注有显著变化的脑区。开发了一种机器学习模型，利用这些确定脑区的平均灌注来预测神经发育结局。与对照组相比，HIE/MRI⊖组在前放射冠、尾状核、额上回、中央前回的PASL灌注显著降低。与HIE/MRI⊖组相比，HIE/MRI⊕组检测到脑灌注有显著增加和降低的变化。三组在小脑、脑干以及丘脑、壳核和苍白球的深部结构中灌注无显著差异。机器学习模型在预测HIE/MRI⊖患者的语言（r = 0.48）和运动（r = 0.57）结局以及预测HIE/MRI⊖和HIE/MRI⊕联合的另一组患者的语言（r = 0.76）和运动（r = 0.53）结局方面显示出显著相关性（P < 0.05）。灌注MRI在检测HIE方面可发挥重要作用，无论常规MRI结果如何，并且可预测HIE幸存者的语言和运动结局。灌注变化还可能揭示再灌注反应和内在自动调节机制的重要见解。我们的结果表明，在临床实践中评估HIE时，灌注成像可能是常规MRI的有用辅助手段。
