Cognitive Domains Classification Using fNIRS-EEG

抽象的。功能性近红外光谱（fNIRS）是一种神经成像技术，其使用低水平的 光（650-900 nm），以非侵入性地测量脑血容量和氧合的变化。同时， 脑电图（EEG）测量神经振荡活动，其可分为各种类型。 节奏频率带。我们提出了一种范式转换分析方法，而不是使用脑电图或 为了估计和测试特定任务的从头大脑激活模式，我们利用现有的 从功能磁共振成像中完成的类似任务中了解大脑活动的常见模式，以创建一组可测试的[null] 假设;即“fNIRS/EEG测量的模式[不]与预期一致（例如） 工作记忆任务的一般分类？” 我们的创新建议是将大脑信号映射到复杂大脑的“认知域”上 功能，而不是任务本身。具体来说，我们的目的是使用一个投影（重新参数化）， 使用由DSM-V定义的六个关键认知领域定义的空间图的大脑信号;语言， 知觉运动、执行功能、复杂注意力、社会认知以及学习和记忆。这些 地图是从数百个现有的功能磁共振成像研究汇总的所有任务的荟萃分析中产生的， 认知领域作为一个关键词，使我们能够从一个大型的各种知识库中开发模型训练， 任务这种将大脑信号重新参数化和压缩到该低维空间中的方法提供了 任务不可知的、统计上可测试的和可解释的信号，可以实时检查。主要 这项研究的目的是： ·目标1。开发和验证认知领域的fNIRS-EEG特征。 ·目标2。fNIRS-EEG特征的重测信度。 目标3。预测模型的定量敏感性-特异性报告。 目标4。实现实时fNIRS-EEG的认知领域分类。

Abstract. Functional near-infrared spectroscopy (fNIRS) is a neuroimaging technique that uses low-levels of light (650-900 nm) to noninvasively measure changes in cerebral blood volume and oxygenation. Meanwhile, electroencephalography (EEG) measures the neural oscillatory activity which can be divided into various rhythms frequency bands. We propose a paradigm shifting analysis approach where instead of using EEG or fNIRS to estimate and test the de novo brain activation pattern for a specific task, we utilize our existing knowledge of common patterns of brain activity from similar tasks done in fMRI to create a set of testable [null] hypotheses; namely, “is the pattern of fNIRS/EEG measured [not] consistent with what is expected from (e.g.) the general category of working memory tasks?” Our innovative proposal is to map brain signals onto the “cognitive domains” underlying complex brain functions rather than task themselves. Specifically, we purpose to use a projection (reparameterization) of brain signals using spatial maps defined by the six key cognitive domains defined by the DSM-V; language, perceptual-motor, executive function, complex attention, social cognition, and learning and memory. These maps are generated from meta-analysis of hundreds of existing fMRI studies pooled across all tasks related to the cognitive domain as a keyword allowing us to develop model training from a large repository of varied tasks. This reparameterization and compression of the brain signals into this low dimensional space provides a task-agnostic, statistically testable, and interpretable signal, which can be examined in real-time. The main aims of this study are: · Aim 1. Develop and validate fNIRS-EEG features in cognitive domains. · Aim 2. Test-retest reliability of fNIRS-EEG features. · Aim 3. Quantitative sensitivity-specificity report of the prediction model. · Aim 4. Implementation real-time fNIRS-EEG for cognitive domain classification.