Over the last 15 years, network analysis approaches based on MR data have allowed a renewed understanding of the relationship between brain function architecture and consciousness. Application of this approach to Disorders of Consciousness (DOC) highlights the relationship between specific aspects of network topology and levels of consciousness. Nonetheless, such applications do not acknowledge that DOC patients present with a dramatic level of heterogeneity in structural connectivity (SC) across groups (e.g., etiology, diagnostic categories) and within individual patients (e.g., over time), which possibly affects the level and quality of functional connectivity (FC) patterns that can be expressed. In addition, it is rarely acknowledged that the most frequently employed outcome metrics in the study of brain connectivity (e.g., degree distribution, inter- or intra-resting state network connectivity, and clustering coefficient) are interrelated and cannot be assumed to be independent of each other. We present empirical data showing that, when the two points above are not taken into consideration with an appropriate analytic model, it can lead to a misinterpretation of the role of each outcome metric in the graph's structure and thus misinterpretation of FC results. We show that failing to account for either SC or the inter-relation between outcome measures can lead to inflated false positives (FP) and/or false negatives (FN) in inter- or intra-resting state network connectivity results (defined, respectively, as a positive or negative result in network connectivity that is present when not accounting for SC and/or outcome measure inter-relation, but becomes not significant when accounting for all variables). Overall, we find that unconscious patients have lower rates of FP and FN for within cortical connectivity, lower rates of FN for cortico-subcortical connectivity, and lower rates of FP for within subcortical connectivity. These lower rates in unconscious patients may reflect differences in their triadic closure and SC metrics, which bias the interpretations of the inter- or intra-resting state network connectivity if the SC metrics and triadic closure are not modeled. We suggest that future studies of functional connectivity in DOC patients (i) incorporate where possible SC metrics and (ii) properly account for the intercorrelated nature of outcome variables.

中文翻译：

---

TBI 患者功能网络分析中结构变化的解释


在过去的 15 年里，基于 MR 数据的网络分析方法使人们能够重新认识大脑功能结构与意识之间的关系。这种方法在意识障碍（DOC）中的应用突出了网络拓扑的特定方面与意识水平之间的关系。尽管如此，此类应用并未承认 DOC 患者在不同组别（例如，病因学、诊断类别）和个体患者内部（例如，随着时间的推移）在结构连接性（SC）方面表现出显着的异质性水平，这可能会影响水平和可表达的功能连接（FC）模式的质量。此外，很少有人认识到，大脑连接性研究中最常用的结果指标（例如，程度分布、静息状态间或静息状态内网络连接性和聚类系数）是相互关联的，不能假设它们独立于彼此。我们提供的经验数据表明，当没有使用适当的分析模型考虑上述两点时，可能会导致对图表结构中每个结果指标的作用的误解，从而导致 FC 结果的误解。我们表明，未能考虑 SC 或结果测量之间的相互关系可能会导致静息状态网络连接结果之间或内部的假阳性 (FP) 和/或假阴性 (FN) 夸大（分别定义为作为网络连接的积极或消极结果，在不考虑 SC 和/或结果测量相互关系时存在，但在考虑所有变量时变得不显着）。 总体而言，我们发现意识不清的患者皮质内连接的 FP 和 FN 率较低，皮质-皮质下连接的 FN 率较低，皮质下连接的 FP 率较低。昏迷患者中较低的比率可能反映了他们的三元闭合和 SC 指标的差异，如果 SC 指标和三元闭合没有建模，这会导致对静息状态间或静息状态内网络连接的解释产生偏差。我们建议未来对 DOC 患者功能连接的研究 (i) 尽可能纳入 SC 指标，并且 (ii) 正确解释结果变量的相互关联性质。