Department of Psychiatry and Behavioral Sciences, Stanford University Medical Center, Stanford, California; Veterans Affairs Palo Alto Healthcare System, and the Sierra Pacific Mental Illness, Research, Education, and Clinical Center, Palo Alto, California; Wu Tsai Neuroscience Institute, Stanford, California.
Department of Psychiatry and Behavioral Sciences, Stanford University Medical Center, Stanford, California; Veterans Affairs Palo Alto Healthcare System, and the Sierra Pacific Mental Illness, Research, Education, and Clinical Center, Palo Alto, California; Wu Tsai Neuroscience Institute, Stanford, California; Department of Clinical Neurophysiology, HUS Diagnostic Center, Clinical Neurosciences, Helsinki University Hospital and University of Helsinki, Helsinki, Finland.
Biol Psychiatry Cogn Neurosci Neuroimaging. 2023 Aug;8(8):805-814. doi: 10.1016/j.bpsc.2022.12.005. Epub 2022 Dec 17.
Noninvasive brain stimulation and neuroimaging have revolutionized human neuroscience with a multitude of applications, including diagnostic subtyping, treatment optimization, and relapse prediction. It is therefore particularly relevant to identify robust and clinically valuable brain biomarkers linking symptoms to their underlying neural mechanisms. Brain biomarkers must be reproducible (i.e., have internal reliability) across similar experiments within a laboratory and be generalizable (i.e., have external reliability) across experimental setups, laboratories, brain regions, and disease states. However, reliability (internal and external) is not alone sufficient; biomarkers also must have validity. Validity describes closeness to a true measure of the underlying neural signal or disease state. We propose that these metrics, reliability and validity, should be evaluated and optimized before any biomarker is used to inform treatment decisions. Here, we discuss these metrics with respect to causal brain connectivity biomarkers from coupling transcranial magnetic stimulation (TMS) with electroencephalography (EEG). We discuss controversies around TMS-EEG stemming from the multiple large off-target components (noise) and relatively weak genuine brain responses (signal), as is unfortunately often the case in noninvasive human neuroscience. We review the current state of TMS-EEG recordings, which consist of a mix of reliable noise and unreliable signal. We describe methods for evaluating TMS-EEG biomarkers, including how to assess internal and external reliability across facilities, cognitive states, brain networks, and disorders and how to validate these biomarkers using invasive neural recordings or treatment response. We provide recommendations to increase reliability and validity, discuss lessons learned, and suggest future directions for the field.
非侵入性脑刺激和神经影像学技术已经在人类神经科学领域引发了一场革命，拥有多种应用，包括诊断亚型分类、治疗优化和复发预测。因此，确定将症状与其潜在神经机制联系起来的稳健且具有临床价值的大脑生物标志物尤为重要。大脑生物标志物必须在实验室内部的类似实验中具有可重复性（即具有内部可靠性），并且在实验设置、实验室、脑区和疾病状态之间具有可推广性（即具有外部可靠性）。然而，可靠性（内部和外部）本身并不足够；生物标志物还必须具有有效性。有效性描述了与潜在神经信号或疾病状态的真实测量值的接近程度。我们提出，在将任何生物标志物用于告知治疗决策之前，应评估和优化这些可靠性和有效性指标。在这里，我们将讨论这些指标，涉及到使用经颅磁刺激（TMS）与脑电图（EEG）相结合的因果脑连接生物标志物。我们讨论了源于多个大型非靶标成分（噪声）和相对较弱的真实大脑反应（信号）的 TMS-EEG 争议，不幸的是，这种情况在非侵入性人类神经科学中经常发生。我们回顾了 TMS-EEG 记录的现状，其中包括可靠的噪声和不可靠的信号的混合。我们描述了评估 TMS-EEG 生物标志物的方法，包括如何在设施、认知状态、脑网络和疾病之间评估内部和外部可靠性，以及如何使用侵入性神经记录或治疗反应来验证这些生物标志物。我们提供了提高可靠性和有效性的建议，讨论了经验教训，并为该领域提出了未来的方向。
