Characterising abnormal early brain development and links to neuropsychiatric disorders using diffusion MRI and machine learning

新出现的证据表明，各种神经精神疾病，无论发病年龄大小，都可能源于早期发育途径的中断(1)。例如，对发育中的人脑的转录分析发现，与这些疾病显著相关的基因在胎儿大脑皮层高表达(2)。由于脑的可塑性是一种持久和持续的特征，在出生后不久活动最高(3)，认知受损的成年人的神经元回路表达可能是由于大脑发育敏感期的损伤造成的。事实上，对人生第一个和九十个年龄段的人进行的皮质地形图比较表明，成熟变化的轨迹可以在不同地区以不同的速度变化(4)。除了风险基因外，围产期发现的环境因素，包括母亲压力、病毒感染、营养不良和产科并发症，也可能导致精神分裂症和双相情感障碍出现行为异常的风险(5，6)。因此，识别大脑在遗传和环境扰动下高度脆弱的关键窗口，可以为潜在的治疗目标和对后来的认知缺陷的早期干预提供新的见解。然而，由于目前关于人脑成熟的大多数知识主要是从死后和动物模型推断出来的，在将这些发现翻译到人类受试者身上时，仍然存在临床无效的可能性。随着非侵入性成像技术的进步，现在可以在神经精神疾病发生之前很久就对新生儿和幼儿的发育过程进行详细的检查。早产通常以非常低的出生体重为特征，是指那些在胎龄不到37周的情况下出生的婴儿；虽然在发达国家，早产只占所有分娩的5%-12%，但他们几乎占所有围产期死亡的75%，与足月出生的同龄人相比，他们患神经发育障碍的风险高4倍，包括精神分裂症、注意力缺陷/多动障碍、自闭症谱系障碍和情绪障碍(7，8)。事实上，流行病学研究表明，随着出生体重和GA的减少，精神疾病发病率会增加(9)。此外，由于早产的许多风险决定因素，包括母亲感染、炎症、应激和出血，与前面提到的神经发育障碍的环境风险因素重叠，早产儿提供了一个有吸引力的模型，以弥合动物和人类研究之间的神经生物学知识差距。因此，检查早产儿大脑结构和功能异常的病因学可以提供对精神疾病潜在治疗窗口的更多洞察。通过利用神经生物学知识和机器学习方法，本研究旨在通过弥散磁共振找出与精神疾病的遗传风险因素和早产儿长期预后相关的神经成像标记物。这项工作的结果可以为围产期脑损伤的神经发育后果提供进一步的洞察，弥合动物和人类研究之间的差距，并有助于在早期关键窗口进行新的治疗干预。为了实现这一目标，本研究将尝试实现以下目标：确定与早产儿神经精神疾病相关的潜在神经发育途径。建立多个变量(遗传变量)与多个变量(d-MRI标记)之间微小关联的统计推断框架。使用现有的新生儿数据集在已识别的路径上执行此类管道的验证。

Emerging evidences suggest that a wide range of neuropsychiatric diseases, irrespective of their age of onset, could arise from the disruptions of early developmental pathways (1). For example, transcriptomic analyses of developing human brain have found genes significantly associated with the disorders to express highly across the foetal cerebral cortex (2). Since the brain plasticity is a long-lasting and continuous feature with the highest activity soon after birth (3), the neuronal circuits expressed in the cognitively impaired adults could result from damages during sensitive periods of the brain development. Indeed, cortical mapping comparison of individuals between the first and ninth decades of life has demonstrated the trajectories of maturational changes can vary across regions and at different rates (4). In addition to the risk genes, environmental factors found during the perinatal period including maternal stress, viral infection, malnutrition, and obstetric complications can also contribute to the risk of behavioural abnormalities seen in schizophrenia and bipolar disorders (5, 6). Thus, identification of critical windows, during which the brain is highly vulnerable to both genetic and environmental perturbations, could provide novel insights to potential therapeutic targets and early intervention to later cognitive deficits. However, since most of the current knowledge on the human brain maturation is extrapolated primarily from post-mortem and animal models, there remains a possibility of clinical ineffectiveness when translating such findings in human subjects. With advancement of non-invasive imaging techniques, it is now feasible to examine the developmental processes with great details in neonates and young children long before the onset of the neuropsychiatric diseases. Preterm births, often characterised by very low birth weights, are those that take place at less than 37 weeks' gestational age (GA); although accounting for only 5-12% of all deliveries in the developed countries, they make up almost 75% of all perinatal deaths and have 4-fold higher risks of developing neurodevelopmental disorders including schizophrenia, attention deficit/hyperactivity disorder, autism spectrum disorder and emotional disorders compared with the term-born peers (7, 8). Indeed, epidemiologic studies have suggested the occurrence of psychiatric morbidity to increase as birthweight and GA decrease (9). Additionally, since many of the risk determinants of preterm labour including maternal infection, inflammation, stress and haemorrhage overlap with previously mentioned environmental risk factors for neurodevelopmental disorders, preterm neonates present attractive model to bridge the gap in neurobiological knowledge between animal and human studies. Therefore, examining aetiology of structural and functional abnormalities in the premature brain could provide more insight into the potential therapeutic windows of psychiatric diseases.By leveraging neurobiological knowledge and machine learning approaches, the study aims to identify neuroimaging markers derived from diffusion MRI that could associate with genetic risk factors for psychiatric diseases and long-term outcome in preterm neonates. The result from this work could provide further insight into neurodevelopmental consequence of perinatal brain injury, bridge the gap between animal and human research and contribute to novel therapeutic interventions in early critical windows. To achieve this aim, the study will attempt to carry out the following objectives: Identify potential neurodevelopmental pathways that correlate with neuropsychiatric diseases in preterm neonates. Build a framework for statistical inference of small associations between many variables (genetic variants) to many variables (d-MRI markers). Perform validation of such pipeline on the identified pathways using existing neonatal dataset.