MRI and Deep Learning for Early Prediction of Neurodevelopmental Deficits in Very Preterm Infants

项目摘要/摘要 在美国，每年大约有100,000名极早产儿(VPI；≤，胎龄32周)出生。 高达35%的人患有明显的神经发育缺陷，从而增加了他们受教育程度较低的风险， 健康和社会成果。不幸的是，目前还不能可靠地诊断出神经发育缺陷。 直到3至5岁。迫在眉睫的挑战在于及早识别出更有可能患上 发展以后的赤字。磁共振成像(MRI)和深度学习(DL)的进展提供了手段 来应对这一挑战。DL在婴幼儿脑MRI数据中的应用可为早期研究开辟新的窗口 预测高危婴儿的神经发育结果并促进其精确化 医药。我们的目标是将DL应用于在足月相同年龄获得的MRI，以早期预测 VPI患者2岁时出现神经发育缺陷(认知、语言和运动)。我们的团队已经确定了三个关键 未来神经发育的准确预测模型所需的组件。1)的DL分析 源自结构磁共振(SMRI)的解剖特征允许检测脑结构异常和 组织病理学；2)来自静息状态功能磁共振成像(rs-fMRI)和 弥散磁共振成像(DMRI)提供对非典型脑连接模式的洞察；以及3)解剖结构的整合 和连接性特征，从而增强了神经发育异常的预测。在这个项目中，我们将 致力于实现以下具体目标。在目标1和目标2中，我们将开发深度分析 和Deep Conn模型独立分析解剖和连通性特征以预测不利因素 神经发育结果。通过解码每个模型，我们将识别、验证和传播一系列 对研究界最具鉴别力的解剖学和连接性特征。在《目标3》中，我们将 开发一个分析解剖和连通性特征的集成DeepAnaConn模型，以及 临床危险因素，用于神经发育缺陷的早期预测。这一模型将帮助临床医生预测 高危早产儿在新生儿重症监护病房首次出院前的近期结局。 我们将使用内部和独立的外部数据来验证模型，并将打开 以协助解释影像和临床表现。我们开发的技术有望得到改进 在医学诊断/预测中的模型保真度与数字逻辑一样，给其他领域带来了革命性的变化。 我们开发的DL模型不仅有利于VPI神经发育缺陷的早期发现，而且 可能使患有其他神经发育和神经疾病的个人受益。这项研究将具有重要的意义 影响公共健康，因为它将允许临床医生针对临床和实验干预疗法 高危婴儿在最佳神经可塑性时期，从而最终改善医疗结果 和耐心的幸福。

Project Summary/Abstract About 100,000 very preterm infants (VPI; ≤32 weeks gestational age) are born every year in the United States. Up to 35% develop noteworthy neurodevelopmental deficits, thereby increasing their risk for poor educational, health, and social outcomes. Unfortunately, neurodevelopmental deficits cannot currently be reliably diagnosed until 3 to 5 years of age. The imminent challenge lies in early identification of infants that are more likely to develop later deficits. Advances in magnetic resonance imaging (MRI) and deep learning (DL) provide means to address this challenge. Application of DL to infant brain MRI data can open up new windows into early prediction of neurodevelopmental outcomes in at-risk infants and facilitate the move towards precision medicine. Our objective is to apply DL to MRI acquired at term equivalent age for early prediction of neurodevelopment deficits (cognitive, language, and motor) at age 2 in VPI. Our group has identified three key components necessary for accurate prognostic models of later neurodevelopment. DL analysis of 1) anatomical features derived from structural MRI (sMRI) allowing detection of brain structural abnormalities and tissue pathologies; 2) brain connectivity features derived from resting-state functional MRI (rs-fMRI) and diffusion MRI (dMRI) giving insights into atypical brain connectivity patterns; and 3) integration of anatomical and connectivity features, thus enhancing abnormal neurodevelopment prediction. In this project, we will dedicate our efforts in accomplishing the following specific aims. In Aim 1 and Aim 2, we will develop deepAna and deepConn models analyzing anatomical and connectivity features independently to predict adverse neurodevelopmental outcomes. By decoding each model, we will identify, validate and disseminate a series of the most discriminative anatomical and connectivity features to the research community. In Aim 3, we will develop an ensemble deepAnaConn model analyzing both anatomical and connectivity features, together with clinical risk factors, for early prediction of neurodevelopmental deficits. This model will help clinicians to predict later outcomes for those at-risk prematurely born infants before initial neonatal intensive care unit discharge. We will validate the models using both internal and independent external data and will open the ‘black-box’ of DL to aid interpretation of imaging and clinical findings. The techniques we develop are expected to improve the modelling fidelity in medical diagnosis/ prognosis in the same way as DL has revolutionized other fields. The DL models we develop will not only benefit early detection of neurodevelopmental deficits in VPI, but also likely benefit individuals with other neurodevelopmental and neurological diseases. This study will significantly impact public health because it will allow clinicians to target clinical and experimental intervention therapies to the most at-risk infants during periods of optimal neuroplasticity, and thus ultimately improve medical outcomes and patient well-being.