An Automated System for Localizing the Effects of Transcranial Magnetic Stimulation

经颅磁刺激(TMS)是一种非侵入性的技术，可以调节人脑的运动和认知功能，并阐明结构与功能的关系。然而，观察到的效果和临床结果有相当大的差异性，这可能归因于个体间功能解剖差异和神经元网络不同反应行为的复杂相互作用。关于TMS如何影响神经回路动力学的精确方式，已经进行了广泛的实验和计算研究。建立和验证这类模型的一个关键前提是知道哪些大脑皮层位置受到有效刺激。然而，仍然很难确定作为观察到的生理或行为影响的基础的神经结构。这严重阻碍了对刺激机制的理解，对诱导效应的解释，以及有效刺激方案的规划。我们最近开发了一个模型框架，并将其应用于TMS诱发的运动诱发电位。这种方法结合了不同线圈位置和/或方向的多个刺激实验，假设在激活部位，电场和运动诱发电位之间的关系在实验条件下是稳定的。我们证明，我们的方法识别出位于脑回冠内的有明显界限的神经结构是运动诱发电位的起源。值得注意的是，我们表明，建议的方法在个人层面上具有很高的歧视能力。在目前的项目中，我们的目标是：(1)我们计划开发一种自动优化线圈位置和方向选择的方法。该方法将仅依赖于个体头部的几何形状，因此能够在实验之前计算出最优刺激方案；(2)我们将把潜在的刺激部位扩展到皮质下白质，并研究白质刺激与实验效果之间的因果关系；(3)我们将把我们的算法扩展到神经网络的识别，从而导致一个大型的组合问题，需要复杂的优化方案。我们的方法将识别外部可观察到的效应与刺激神经种群之间的多变量关系；(4)由于定位技术迄今仅应用于运动诱发电位，我们将把该方法推广到更复杂的大脑过程和实验，并将其应用于更广泛的研究领域；(3)基于这些知识，我们将能够对大脑中的语言区域进行精确和个别的定位；(6)该方法将与机器人系统集成，从而能够自动识别结构与功能的关系，排除潜在的人为误差源。在这方面，我们试图确定这样一个自动化系统的优点和缺点。

Transcranial magnetic stimulation (TMS) is a non-invasive technique to modulate motor and cognitive functions in the human brain and elucidate structure-function relationships. However, there is considerable variability in the observed effects and clinical outcomes, which may be attributed to a complex interplay of interindividual functional-anatomical differences and the variable Response behaviour of neuronal networks. There has been extensive experimental and computational research on the precise way howTMS affects the dynamics of neural circuits. A crucial prerequisite for building and validating such models is to know which cortical locationsare effectively stimulated. Yet, it is still difficult to identify the neural structures that underlie the observed physiological or behavioural effects. This is a serious obstacle to the understanding of the mechanism of stimulation, the interpretation of the induced effects,and the planning of effective stimulation protocols. We recently developed a modelling framework, which we applied to TMS induced motor evoked potentials. This approach combines multiple Stimulation experiments with different coil positions and/or orientations, assumingthat at the site of activation the relationship between electric field and motor evoked potentials is stable across experimental conditions. We demonstrated that our method identifies sharply bounded neural structures located in the gyral crowns as origin of the motor evokedpotentials. Notably, we showed that the proposed approach has high discriminatory power on the individual level. In the current project we aim at the following objectives: (1) we plan to develop a method to automatically optimize the choice of coil positions and orientations.This method will only rely on the individual head geometry and therefore enables the computation of the optimal stimulation scheme prior to the experiment; (2) we will extend potential stimulation sites to subcortical white matter and investigate causal relationships between white matter stimulation and experimental effects; (3) we will extend our algorithm to the identification of networks of neural populations, leading to a large combinatorial problem that requires sophisticated optimization schemes. Our method will identify multivariaterelationships between externally observable effects and stimulation of neural populations; (4) as so far localization techniques have only been applied to motor evoked potentials, we will generalize the method to more complex brain processes and experimentalparadigms in the sensorimotor and language domains; (5) with this knowledge, we will be able to precisely and individually localize language regions in the brain; (6) the method will be integrated with a robotic system, thus allowing for an automated identification of structure-function relationships, excluding potential human error sources. In this line, we seek to establish the benefits and drawbacks of such an automated system.