经颅磁刺激（Transcranial magnetic stimulation，TMS）是一种安全的、无创的神经调控技术，已广泛地应用于神经相关疾病的临床检测和治疗。将TMS技术与脑电（Electroencephalogram，EEG）技术同步结合，形成一种新的脑成像工具，称之为TMS-EEG。TMS-EEG在非侵入性探测大脑结构与功能关系上具有独特的优势，能够无创检测大脑的兴奋性、连接性和可塑性。前期研究发现，TMS的刺激效果受脑状态影响，降低了TMS-EEG检测的可重复性。通过实时锁定特定脑状态下TMS，可以保证TMS-EEG成像效果的一致性，发展实时闭环TMS-EEG技术成为TMS重要研究方向。另一方面，在认知加工过程中进行TMS-EEG成像，可以呈现任务相关的神经连接，但其成像效果受刺激时刻影响，如何精确地探测出与任务相关的神经网络，也是一个重要的问题。围绕上述问题，本论文研制了有效的TMS-EEG装置和闭环TMS系统，并探索了锁相闭环TMS-EEG生理特征和任务相关TMS-EEG成像，取得主要成果如下：

1、优化和研制了有效的TMS-EEG装置和闭环TMS系统

首先，优化设计了TMS开关电源、刺激线圈、放大器的模拟电路与电极，降低了TMS装置的电磁噪声和线圈噪音并提高了EEG放大器的抗TMS伪迹的能力，搭建一套TMS-EEG装置。其次，通过在6名正常人被试者和13名意识障碍患者开展了TMS-EEG实验，计算TMS诱发电位（TMS-evoked potential，TEP）和快速扰动复杂指数（Fast perturbational complexity index，PCIst），验证了研制的TMS-EEG装置的有效性。最后，设计并讨论了离散傅里叶变换、最小二乘估计的自回归模型、长短时记忆网络和koopman预测等四种预测控制方法，并根据这些方法的预测表现，建立了基于自回归模型的预测控制模型，实现了实时精准的闭环TMS控制。

2、开展了锁相闭环TMS-EEG神经特征提取研究

首先，以枕叶alpha相位作为控制量，以左侧初级运动皮层为刺激靶区开展了不同枕叶alpha相位诱导的锁相闭环TMS-EEG实验，收集了30名健康被试者（年龄：25.23±2.74岁）的锁相闭环TMS-EEG数据。其次，验证了闭环TMS-EEG系统预测控制准确性和分析TMS-EEG试次间相位一致性（Inter trial phase coherence，ITPC），评估了锁相闭环TMS系统的效能：该系统能实现有效的闭环控制，且闭环TMS-EEG的试次间一致性显著高于开环TMS-EEG。最后，计算了TMS-EEG数据的TEP、诱发神经振荡、连接性、脑网络和PCIst指数等指标，综合地提取了锁相闭环TMS-EEG时域、频域和空间域神经特征。分析结果发现：N45-P70-N100 TEP成分波形及其地形分布受到TMS刺激时刻枕叶alpha相位的显著影响；gamma网络连接性显著依赖于枕叶alpha相位，180度相位条件下的多个通道的连接性显著高于0度相位条件下；在0度相位条件下，gamma网络具有最低的聚类系数（F_{2.079, 54.055}=7.592，p=0.001）、最长的最短路径长度（F_{2.163, 56.237}=7.638，p=0.001）和最大的小世界属性值（F_{2.422, 62.972}=6.195，p=0.020），而在180度相位条件下脑网络结果则相反；四种相位条件刺激下PCIst值没有显著差异（F_{2.538, 65.987}=2.231，p=0.103），但在90度相位条件下刺激时，PCIst的试次间变异系数更小（F_{2.793, 72.618}=3.275，p=0.029）。

3、开展了眼跳任务下TMS-EEG实验研究

首先，招募了30名被试者（年龄：22.97±2.27）完成眼跳任务实验。在眼跳目标出现前100ms、前50ms、出现时、出现后50ms和出现后100ms分别在额叶眼动区（Frontal eye field，FEF）和顶叶皮质（Posterior parietal cortex, PPC）脑区进行TMS刺激，同时收集眼跳实验过程中各个时刻的TMS-EEG数据。其次，以有向传递函数（Directed transfer function, DTF）方法计算alpha、beta和gamma频段下各个脑电通道之间的信息流向，探究了眼跳任务过程中TMS刺激时刻对任务下TMS-EEG相关神经连接成像效果的影响。结果发现：目标出现前100ms、前50ms的FEF和PPC相关神经网络的各频段的信息流较少，目标出现时信息流数量开始增多；beta频段和gamma频段目标出现后50ms信息流数量达到峰值，并在出现后100ms下降；alpha频段在目标出现时、出现后50ms、100ms的信息流数量没有差异。最后，将各时刻、各频段的信息流数据进行汇总，构建了眼跳任务相关的TMS-EEG功能时序图谱，呈现了任务相关的神经网络连接的动态变化。

综上所述，本文设计并构建了闭环TMS-EEG系统，分别研究了TMS-EEG时刻依赖性问题：TMS-EEG对锁定脑电特征的时刻依赖性和TMS-EEG对锁定任务加工过程的时刻依赖性，开展锁相闭环TMS-EEG研究和认知任务下TMS-EEG研究。期望基于本研究结果，可以在未来开展更加精准和个性化的TMS-EEG科研和临床应用。

Transcranial magnetic stimulation (TMS) is a safe and non-invasive brain stimulation technology, which has been widely used in the clinical treatment of neuro-related diseases. By combining TMS technology with synchronous electroencephalogram (EEG) technology, a new imaging tool, TMS-EEG plays a unique advantage in non-invasive detection of brain function and structure. Because the stimulation effect of TMS is affected by the nerve state at the time of TMS stimulation, the repeatability of TMS-EEG experiment needs to be improved. Closed-loop TMS-EEG technology can reduce the variability of TMS-EEG results by locking the brain state and ensuring the consistency of TMS stimulation. By performing TMS-EEG imaging when the subjects complete the relevant cognitive processing tasks, the TMS stimulation signal can be used as a "probe" to detect the task-related neural pathways. However, when TMS stimulation is performed during the task completion determines whether the activated neural pathways can be found effectively. Improving the repeatability of TMS-EEG experimental results based on closed-loop TMS-EEG and improving the time accuracy of task-related TMS-EEG imaging technology constitute the time selection problem of TMS-EEG. Based on the above issues, this dissertation has developed an effective TMS-EEG device and closed-loop TMS system, and explored the physiological characteristics of phase-locked closed-loop TMS-EEG and task related TMS-EEG imaging. The main achievements are as follows:

(1) First, in order to meet the platform requirements of TMS-EEG time selection research, by reducing the electromagnetic noise and coil noise of TMS device and improving the ability of EEG amplifier to resist TMS artifact, this research has built a reliable TMS-EEG device, which can effectively record the TMS-EEG data and achieve synchronous trigger function with the host computer. TMS-EEG experiment was carried out in 6 health subjects and 13 patients with consciousness disorders, and the reliability of the system was verified by calculating TMS-evoked potential (TEP) and fast perturbation complexity index (PCIst). A closed loop TMS-EEG system is built based on the matlab GUI platform. Four methods such as discrete Fourier transform, autoregressive model of least squares estimation, long short-term memory networks and koopman prediction are compared in resting EEG data. The autoregressive model method is finally selected as the optimal scheme for predictive control, which can take into account the real-time closed-loop requirements of prediction accuracy and time complexity.

(2) Then, the phase-locked closed-loop TMS-EEG experiment induced by different occipital alpha phases was carried out with the occipital lobe alpha phase as the control quantity, 100% resting motor threshold (RMT) as the stimulation intensity, 3 – 4s as the stimulation interval, and the left primary motor cortex as the stimulation target area. By collecting the phase-locked closed-loop TMS-EEG data of 30 healthy subjects (age: 25.23 ± 2.74 years), the predictive control accuracy of the closed-loop TMS-EEG system was verified, and the characteristics of the TMS-EEG results such as inter-trial phase coherence (ITPC), TEP, evoked neural oscillation, connectivity, brain network and PCIst index were analyzed. The results show that the ITPC value of closed-loop TMS-EEG is significantly higher than that of open-loop TMS-EEG; N45-P70-N100 TEP component waveform and its topographic distribution were significantly affected by the occipital lobe alpha phase at the time of TMS stimulation; Gamma network connectivity is significantly dependent on the occipital lobe alpha phase, and the connectivity of multiple channels under 180 phase condition is significantly higher than that under 0 phase condition; Under the condition of 0 phase, the gamma network has the lowest clustering coefficient (F_{2.079, 54.055}=7.592, p=0.001), the longest shortest path length (F_{2.163, 56.237}=7.638, p=0.001) and the largest small-world attribute value (F_{2.422, 62.972}=6.195, p=0.020), while under the condition of 180 degree phase, the results of brain network are opposite; There was no significant difference in the value of PCIst under the four phase conditions (F_{2.538, 65.987}=2.231, p=0.103), but the coefficient of variation of PCIst between tests was smaller under the 90 degree phase conditions (F_{2.793, 72.618}=3.275, p=0.029).

(3) Finally, in order to explore the effect of TMS stimulation time during the task on the imaging effect of TMS-EEG related nerve pathway under the task, 30 subjects (age: 22.97 ± 2.27) were recruited to complete the saccade task experiment. TMS stimulation was performed in the frontal eye field (FEF) and the parietal cortex (PPC) brain areas 100 ms before, 50 ms before, at, 50 ms after and 100 ms after the occurrence of the saccade target, and TMS-EEG data at each time during the saccade experiment were collected. The directed transfer function (DTF) method is used to calculate the information flow between EEG channels in alpha, beta and gamma frequency bands. The information flow in each frequency band of FEF and PPC-related neural pathways 100 ms and 50 ms before the target appears is less, and the number of information flow begins to increase when the target appears; The number of information flows in the beta band and gamma band reaches the peak 50 ms after the target appears, and decreases 100 ms after the target appears; There is no difference in the number of information flows in the alpha frequency band at the time of target occurrence, 50ms and 100ms after the occurrence. Summarize the information flow data of each time and frequency band, and construct the TMS-EEG functional time series map related to the saccade task, which can show the dynamic changes of the task-related nerve conduction pathway.

To sum up, by establishing a closed-loop TMS-EEG system, this thesis aims at two aspects of TMS-EEG time selectivity: how to lock the neural activity state at the time of stimulation and how to improve the time resolution of neural pathway detection, and carries out phase-locked closed-loop TMS-EEG research and task TMS-EEG research. It is expected that based on the results of this study, more accurate and personalized scientific research and clinical applications can be carried out in the future.