Multivoxel pattern analysis of functional brain imaging data for the prediction of the development and maintenance of alcohol use disorders

以前的神经影像学研究旨在预测酒精成瘾的临床结果提供了令人鼓舞的结果。然而，传统的单变量统计方法存在这样的问题，即，尽管存在统计学上显著的关系，但单个替代标志物（例如给定脑区域中的活动）的预测值通常较低。多体素模式分析（MVPA）可以通过跨多个体素累积信息来显著提高预测能力，即，使用包含在大脑活动的分布式模式中的信息，而不是单个大脑区域。MVPA已成功应用于个体受试者水平和受试者之间的诊断或预后分类。来自项目1的功能和结构磁共振成像（MRI）数据将用于预测年轻人的病理性饮酒模式。将根据项目2的MRI数据预测酒精依赖患者的治疗结局。至关重要的是，MVPA将用于在功能和结构MRI扫描中识别这些特征或特征模式，从而可靠地预测未来的临床结果。这些知识将为病理性饮酒的神经生物学机制的发病模型提供信息，并有助于根据个人的神经生物学风险状况制定雅阁干预措施。

Previous neuroimaging studies aimed at predicting clinical outcome in alcohol addiction have provided encouraging results. However, conventional univariate statistical approaches suffer from the problem that the predictive value of single surrogate markers, such as activity in a given brain region, is usually low despite the presence of a statistically significant relationship. Multivoxel pattern analysis (MVPA) can dramatically increase predictive power by accumulating information across multiple voxels, i.e., to use the information contained in a distributed pattern of brain activity rather than a single brain region. MVPA has been successfully applied both at the individual subject level and for diagnostic or prognostic classification between subjects. Functional and structural magnetic resonance imaging (MRI) data from Project 1 will be used to predict pathological drinking patterns in young adults. Treatment outcomes in alcohol dependent patients will be predicted on the basis of MRI data from Project 2. Crucially, MVPA will be used to identify those features, or patterns of features, in functional and structural MRI scans that reliably predict future clinical outcomes. This knowledge will inform pathogenetic models for neurobiological mechanisms underlying pathological drinking and help tailor therapeutic interventions in accord with an individual¿s neurobiological risk profile.