Li Lin, Zeng Li, Lin Zi-Jing, Cazzell Mary, Liu Hanli
Joint Graduate Program between University of Texas at Arlington and University of Texas Southwestern Medical Center, University of Texas at Arlington, Department of Bioengineering, Texas 76019, United States.
University of Texas at Arlington, Department of Industrial and Manufacturing Systems Engineering, Texas 76019, United States.
J Biomed Opt. 2015 May;20(5):50801. doi: 10.1117/1.JBO.20.5.050801.
Test-retest reliability of neuroimaging measurements is an important concern in the investigation of cognitive functions in the human brain. To date, intraclass correlation coefficients (ICCs), originally used in interrater reliability studies in behavioral sciences, have become commonly used metrics in reliability studies on neuroimaging and functional near-infrared spectroscopy (fNIRS). However, as there are six popular forms of ICC, the adequateness of the comprehensive understanding of ICCs will affect how one may appropriately select, use, and interpret ICCs toward a reliability study. We first offer a brief review and tutorial on the statistical rationale of ICCs, including their underlying analysis of variance models and technical definitions, in the context of assessment on intertest reliability. Second, we provide general guidelines on the selection and interpretation of ICCs. Third, we illustrate the proposed approach by using an actual research study to assess interest reliability of fNIRS-based, volumetric diffuse optical tomography of brain activities stimulated by a risk decision-making protocol. Last, special issues that may arise in reliability assessment using ICCs are discussed and solutions are suggested.
神经影像测量的重测信度是人类大脑认知功能研究中的一个重要问题。迄今为止，最初用于行为科学中评分者间信度研究的组内相关系数（ICC），已成为神经影像和功能近红外光谱（fNIRS）信度研究中常用的指标。然而，由于ICC有六种常见形式，对ICC的全面理解是否充分将影响人们在信度研究中如何正确选择、使用和解释ICC。我们首先在重测信度评估的背景下，对ICC的统计原理进行简要回顾和讲解，包括其潜在的方差分析模型和技术定义。其次，我们提供关于ICC选择和解释的一般指南。第三，我们通过一项实际研究来说明所提出的方法，该研究旨在评估基于fNIRS的、由风险决策协议刺激的大脑活动的体积扩散光学断层扫描的兴趣信度。最后，讨论了使用ICC进行信度评估时可能出现的特殊问题并提出了解决方案。
