新生儿脑MRI对痉挛型脑瘫手功能分级的预测研究

Spastic cerebral palsy (SCP) is the most common type of cerebral palsy. Manual ability deficits can be frequently found on SCP patients. These patients lose the chance to learn knowledge through the perception by hands during the early development. The severe ones will lose the self-care ability in daily life. Early interventions based on manual ability levels will improve the prognosis. However, the assessment of manual ability levels would be accomplished typically after the age of 1 year. It is urgently necessary to predict manual ability levels of SCP during the neonatal period. Therefore, this project tries to solve the key problems for mapping the relationship between manual ability levels and brain lesions by using the neonatal magnetic resonance imaging (MRI). Based on the local MRI database, main contents of this project are arranged as follows: 1) Image segmentation will be performed by using the convolutional neural networks to detect brain lesions on neonatal MRI images. 2) This project will analyze structural alterations associated with brain lesions by combing T1 weighted imaging, T2 weighted imaging, and diffusion kurtosis imaging. Registration of brain atlas and fiber-tract quantification will be performed to characterize the spatial distribution of structural alterations. 3) Based on neonatal MRI features and the follow-up outcomes of manual ability levels, multivariate lesion-symptom mapping will be performed to reveal the relationship between manual ability levels and neonatal brain lesions. This project will improve the prediction of manual ability levels of SCP and provide evidences for the choice of early interventions.

痉挛型脑瘫是最常见的脑瘫类型，多伴有手功能障碍，导致患儿在发育的关键时期无法通过手获得足够的感官经验，严重者丧失生活自理能力。早期干预有助于改善预后，手功能分级是选择适宜干预措施的重要依据。然而，手功能分级的评估仍需在1岁以后。在新生儿期预测手功能分级已成为迫切需求，其核心问题是“手功能分级与早期脑损伤的映射关系是什么”。鉴于此，在前期构建的影像数据库基础上，本项目拟1）通过新生儿脑磁共振成像（MRI）,利用卷积神经网络实现病灶的自动化检测；2）联合T1加权、T2加权与扩散峰度成像对脑形态学与微结构属性进行量化分析，并通过脑图谱仿射与空域分析实现脑结构属性改变的精准定位；3）基于新生儿脑MRI特征与随访的手功能分级，使用病灶-症状映射方法揭示痉挛型脑瘫手功能分级与新生儿脑损伤多参量特征的映射关系。本研究有望推进对痉挛型脑瘫手功能分级的预测，为早期干预措施的选择提供依据。

手功能的正常发育是儿童获得感官经验进而掌握精细动作能力的基础，脑瘫患儿多伴有手功能障碍，严重影响患儿的生活自理能力。早期评估与早期干预有助于改善预后。然而，发育早期儿童的行为能力尚未发育完善，利用常规行为学评估方法进行手功能早期评估的难度大，脑影像或有助于手功能早期评估，磁共振成像适宜于儿童脑部检查。但是，手功能障碍与脑损伤的映射关系尚不明确，基于磁共振成像对手功能进行评估面临巨大挑战。.针对上述临床需求与技术挑战，本项目开展了脑部病灶的自动化检测方法研究、脑部结构属性改变的量化评估与空间定位方法研究、脑瘫患儿手功能分级与病灶特征的映射研究等三方面的研究工作。通过项目实施，（1）对新生儿脑损伤病灶检测方法进行了优化设计，建立了双通道神经网络架构，提升了新生儿脑白质损伤病灶区域分割的准确性；（2）基于机器学习实现了脑瘫患儿易损区域的图像分割；（3）基于生成对抗网络构建了不同中心参量图像的同质化方法；（4）基于磁共振成像多参量变化的组合模式建立了脑发育进程的评估方法；（5）基于脑瘫患儿脑网络属性的早期变化特点，并联合脑白质微结构属性量化指标，提升了脑瘫风险预测的灵敏度与特异度；（6）探索了手节区形态类型及其在发育过程中的变化特点，在发育早期发现了一种新的类型：不成熟欧米伽型手节；（7）揭示了痉挛型偏瘫儿童手节区形态的变化特点；（8）通过脑瘫患儿病灶-脑连接-症状映射揭示了脑瘫患儿手功能障碍的关键损伤脑区，并基于影像学评估建立了轻度与重度手功能障碍的鉴别模型。.通过项目实施，所建立的图像处理与分析方法有望为脑发育及其损伤研究提供方法学支撑，所发现的不成熟手节类型有助于理解脑发育规律，所揭示的脑瘫患儿手功能障碍关键损伤脑区有助于为手功能障碍的早期诊断以及脑部干预靶点的选择提供线索。