Brain AnalyzIR: A software platform for improving scientific rigor in functional NIRS statistical analysis

摘要 功能性近红外光谱(FNIRS)是一种非侵入性的神经成像手段，它使用低水平的 光来测量大脑中诱发的血液动力学变化。这种技术越来越受欢迎。 在过去的几十年里，由于它的通用性和便携性，以及这项技术在独特的 实验情况和受试者群体，例如关于儿童、婴儿或使用生态有效的研究 实验设计(散步、社交互动等)。随着该领域终端用户数量的增长， 必须为这些研究的分析和解释建立科学严谨的最佳做法。一个 FNIRS领域的谬误在于直接引入了来自其他模式的方法和解释(例如， 功能磁共振成像)没有对fNIRS特有的噪声和信号特性进行适当的适应和推广 数据的一部分。此外，到目前为止，许多fNIR方法的发展都是基于自组织的 在特定数据集下对这些算法的观察。因此，最终用户经常使用设计的方法 对于与他们自己的数据不匹配的统计假设。未使用适当的统计模型或未满足要求 假设经常导致高的假阳性率和低的科学严谨性，这就是在 许多先前的近红外光谱研究。该生物医学研究小组(BRG-R01)项目的目标是建立 FNIRS分析的当前最佳实践和基于量化的未来发展的基础设施 通过接收器操作员特征分析、偏差量化等进行方法比较。 项目还将建立一个开放源码的fNIRS数据库，以便对 研究fNIRS信号的各种性质，并量化其对统计模型的影响。我们集团有很长的一段时间 过去15年的fNIRS分析和开源软件开发的历史，被认为是 在fNIR分析中排名靠前的实验室。该项目的具体目标是： 目标1.开发开放的fNIRS数据库和基准平台，用于测试和表征 开发新的算法和统计方法。 目的2.确定一般噪声模型和分类噪声模型下fNIRS分析的最佳做法。 目标3.继续开发和改进以终端用户为重点的fNIRS具体分析模型 需求和反馈。 目的：4.方法的传播和培训。

Abstract Functional near-infrared spectroscopy (fNIRS) is a non-invasive neuroimaging modality that uses low-levels of light to measure evoked hemodynamic changes in the brain. This technique has been growing in popularity over the last several decades due its versatility and portability and the applicability of this technique in unique experimental situations and subject populations, such as studies on children, infants, or using ecologically valid experimental designs (walking, social interaction, etc). As the number of end-users in this field grows, it is important to establish scientifically rigorous best practices for analysis and interpretation of these studies. A fallacy of the fNIRS field has been the direct import of methods and interpretations from other modalities (e.g. functional MRI) without proper adaptation and generalization for the fNIRS-specific noise and signal properties of the data. Furthermore, to date, the development of many fNIRS methods has been based on ad-hoc observations of these algorithms under specific datasets. As a result, end-users often use methods designed for statistical assumptions that do not match their own data. Failure to use proper statistical models or unmet assumptions often results in high false-positive rates and poor scientific rigor and this has been the case in many prior fNIRS studies. The goal of this Biomedical Research Group (BRG-R01) project is to establish current best practices for fNIRS analysis and an infrastructure for future development based on quantitative comparisons of methodologies via receiver operator characteristics analysis, quantification of bias, etc. This project will also establish an open-source fNIRS database to allow characterization and classification of the various properties of fNIRS signals and to quantify their effect on statistical models. Our group has a long history of fNIRS analysis and open-source software development over the last 15 years and is considered one of the top labs in fNIRS analysis. The specific aims of this project are: Aim 1. Development of an open fNIRS database and benchmarking platform for testing and characterizing the development of new algorithms and statistical methods. Aim 2. Determination of best practices for fNIRS analysis under general and categorized noise models. Aim 3. Continued development and improvement of fNIRS-specific analysis models with focus on end-user needs and feedback. Aim4. Dissemination and training of methods.