Abstract Brain-state-guided and closed-loop transcranial magnetic stimulation (TMS) protocols have emerged as methods for decreasing the variability and increasing the therapeutic effectiveness of stimulation protocols. However, most existing brain-state-dependent TMS systems only control the timing of stimulation, while the location is fixed and manually adjusted between blocks or sessions. This limits flexible targeting of distributed networks. We developed a system that jointly manages TMS pulse timing and location automatically controlled by an electroencephalography (EEG)-based brain– computer interface (BCI). A machine-learning algorithm infers the brain state in real time to guide the robotic coil placement and target. We present a proof-of-concept study in which a BCI controlled both the target site and the timing of TMS. A pre-trained convolutional neural network discriminated between resting state and movements performed with the right or left hand; the classifier output determined the hemisphere in which primary-motor-cortex hand area was stimulated and when. Preprocessing and decoding of 2-s EEG segments required 150 ms, and the robot took 7.5 s to move from the vertex home position to the predefined motor targets. The EEG-BCI-guided robotic TMS system expands the toolkit for brain-state-dependent and closed-loop neurostimulation by enabling control of stimulus location based on volitional brain activity. Thus, the system can benefit both neuroscience research and clinical neuromodulation applications. A prominent application of the system is automatically controlling spinal cord injury or motor disorder TMS rehabilitation with motor imagery, optimizing stimulation timing to the brain state producing optimal rehabilitation results.
Introduction Transcranial magnetic stimulation (TMS) is widely employed to treat various psychiatric and neurological disorders. However, TMS protocols typically rely on generalizations, particularly in selecting stimulation intensities, leading to suboptimal and variable outcomes. Combining TMS with electroencephalography (EEG) offers a potential solution by allowing direct monitoring of stimulation effects. In this study, we investigate how features of the TMS–EEG signal change with intensity to identify thresholds implying qualitative shifts in the brain response. Methods We stimulated eight subjects at both the primary motor cortex (M1) and the pre-supplementary motor area (pre-SMA) with navigated TMS at 15 closely spaced intensities and measured TMS-evoked EEG responses (TMS-evoked potential, TEP) with 60 trials per intensity. TEP thresholds were identified with three methods: by selecting the intensity where the TEP peak-to-peak exceeds 6 μV, or by fitting either piecewise-linear or sigmoid curves into the power spectral density (PSD) frequency components to identify nonlinear intensity behavior. Results The identified TEP thresholds varied depending on subject, target, and identification method. In M1, the thresholds attained with the fixed-amplitude and piecewise PSD fit methods averaged around the motor threshold, and in pre-SMA around 120% of the motor threshold. The TEP thresholds yielded by the sigmoid fit method were higher in intensity, and least consistent between subjects. Conclusions Our findings support the hypothesis that detectable changes in EEG patterns occur at specific TMS intensities. These results provide a basis for individualized stimulation dosing, potentially enhancing therapeutic efficacy and reliability. Future research should focus on refining these methods and validating their clinical applicability across diverse conditions and patient populations. ### Competing Interest Statement RJI and VHS are inventors on patents and/or patent applications on TMS technology. IG, PL, RJI, and VHS have received consulting fees from Nexstim Plc. unrelated to this study. RJI and VHS are co-founders of Cortisys Ltd. Other authors have no competing interests to declare. European Research Council, 810377 Wellcome Leap Finnish Indian Consortia for Research and Education Swedish Cultural Foundation Finnish Cultural Foundation, https://ror.org/027xav248 Research Council of Finland, 349985
Abstract Speech cortical mapping by means of navigated repetitive transcranial magnetic stimulation (SCM nrTMS) provides neurosurgeons with noninvasive prior information about individual’s cortical speech network. Individualized mapping is required, since the exact locations and activation patterns of speech production show high variability between individuals. We hypothesized that magnetoencephalography (MEG) data of an individual’s speech production could guide the SCM TMS process temporally and spatially, leading to higher error rates at MEG-defined locations with TMS pulse timings coinciding with MEG activity. 13 healthy subjects participated in MEG and TMS measurements, where the timing of the TMS pulse (PTI; picture-to-TMS interval) was adjusted based on the individual’s MEG activation in a picture naming task. At the group level, significant correlations were observed between the latency of the peak MEG activation and the PTI that produced the highest speech error rate. The MEG peak preceded the best PTI by 132 ms (R=0.713, p =0.006) across the entire stimulation area in the lateral left hemisphere, and by 103 ms (R=0.673, p =0.012) in the left frontal regions. We found 17 combinations of PTI and stimulation area in which the subject’s speech error rate increased significantly compared to their average error rate. Our findings suggest that optimal PTIs are highly individual, and that individualizing the PTI according to MEG activation provides a straightforward method for accounting individual variability in speech function and may increase the sensitivity and utility of SCM TMS.
Transcranial magnetic stimulation (TMS) is a non-invasive technique to stimulate the brain, while electroencephalography (EEG) is a non-invasive technique to record its electrical activity. Their combined use (TMS-EEG) has been established only relatively recently, after successful development of TMS-compatible EEG amplifiers. TMS-EEG offers the unparalleled opportunity to directly perturb the brain with TMS and simultaneously record its response with EEG. This allows inferences on causal input-output relationships, therefore going critically beyond purely observational techniques, such as resting-state EEG or functional MRI, in the study of brain dynamics. This consensus review updates the work of Tremblay and coworkers [Clin Neurophysiol 2019; 130: 802-844]. Since then, substantial advances have been made in understanding contamination of TMS-EEG signals by physiological and non-physiological artifacts, as well as in developing strategies to avoid or control them. In parallel, new insights have emerged regarding the physiological mechanisms underlying TMS-EEG responses and their diagnostic and prognostic utility in a broad range of psychiatric and neurological disorders. As such, TMS-EEG is rapidly shaping a dynamic new field in clinical neurophysiology and neuroscience. This review provides a critical and comprehensive synthesis of current knowledge, including practical guidance for implementing TMS-EEG in the clinical setting.
OBJECTIVE:This study assessed the test-retest reliability of TMS-evoked potentials (TEPs) across two cortical regions-dorsolateral prefrontal cortex (DLPFC), and angular gyrus-in comparison to motor cortex (M1), using individualized and literature-based targeting approaches. The study also compared the reliability of different double-pulse TMS protocols. METHODS:Seventeen healthy participants underwent two TMS-EEG sessions spaced by at least one week, with targets for DLPFC and angular gyrus identified using resting-state functional connectivity (RS) and Neurosynth-based functional overlays. RESULTS:M1 demonstrated the highest TEP reliability (Concordance Correlation Coefficient (CCC) mean = 0.59), while DLPFC (CCCmean = 0.40) and angular gyrus (CCCmean = 0.45) showed lower reliability, particularly for anterior DLPFC targets. Neurosynth-based DLPFC targets exhibited slightly higher CCC values (mean CCC = 0.57) compared to RS-based targets (mean CCC = 0.30), but the difference was not statistically significant. No significant differences in reliability were found across TMS protocols. CONCLUSION:While fMRI-based targeting optimizes the engagement of functional networks, TMS-EEG needs to prioritize signal-to-noise ratio (SNR) to minimize artifacts and maximize reliability. SIGNIFICANCE:This work bridges the gap between fMRI-guided targeting and TMS-EEG applications.
Transcranial magnetic stimulation-electroencephalography (TMS-EEG) biomarkers have recently become available as a means to obtain new understanding of the causal chains of neuronal signaling in the brain. This is a key piece in the puzzle of how the brain is organized and how it works. Using dMRI tractography, we can map the circuit beneath a chosen cortical target; TMS can then stimulate it, and EEG records responses that reflect-and may even be caused by-activity in that structural circuit. The chain of events after stimulus delivery can be observed and quantified using current neuroimaging and TMS-EEG technology, a matter of tremendous relevance on how to approach novel therapeutic approaches in clinical conditions. Herein, we elaborate upon a perspective of how groundbreaking multi-locus TMS (mTMS) technology associated with EEG and multimodal neuroimaging can be applied to modulate the flow dynamics of the glymphatic system (GS). The enhancement of the GS waste clearance functionality has been shown to improve significantly symptom severity in neurodegenerative disorders such as Alzheimer's (AD) and Parkinson's disease (PD) or long COVID. In this perspective paper, we consider that next-generation therapeutics using versatile technologies such as noninvasive neuromodulation and neuroimaging will provide important benefits in public health and in how society can address the management of these difficult-to-deal-with ailments more effectively.
Interictal epileptiform discharges (IEDs) are pathological hypersynchronous bursts of electrical brain activity that occur between seizures in patients with epilepsy. IEDs are caused by transient brain states that are difficult to predict, making them a challenging neurophysiological and technological case for brain-state-dependent stimulation. Administering stimulation at IED onset may provide insight into the epileptic network and optimize neurostimulation therapies. Here, we assessed the feasibility of IED-triggered transcranial magnetic stimulation (TMS) in two children with self-limited epilepsy with centrotemporal spikes (SeLECTS), a common pediatric epilepsy in which IEDs emerge from the motor cortex. A convolutional neural network (CNN) was trained on the participants' pre-recorded electroencephalography (EEG) data with IEDs annotated by an epileptologist. The CNN was integrated into an EEG-processing pipeline that classified EEG segments as "IED" or "non-IED" in real time. With this pipeline, TMS pulses were administered during IED or non-IED periods in an interleaved, randomized design. We stimulated both the motor cortex generating the IEDs and the contralateral motor cortex and tested the impact of IEDs on TMS-evoked potentials (TEPs). Our study demonstrated that TMS can be timed to IEDs and that there is a site-specific increase in TEP amplitude when stimulating during IEDs. Out of the TMS pulses aimed at an IED, 39% and 19% were successfully delivered during an IED for the two participants, respectively. For future research, we propose ways to address the methodological challenges of IED-timed TMS, enabling brain-state-dependent TMS for epilepsy research and treatment.
The cerebral cortex is organized into structurally and functionally segregated networks, enabling the human brain to process information highly efficiently. Transcranial magnetic stimulation (TMS), in combination with electroencephalography (EEG), offers a non-invasive approach to probing brain networks, revealing cortical excitability and causal connectivity. However, this method faces two significant challenges: (a) ensuring the quality of TMS-evoked potentials (TEPs) to maximize information gain, often requiring comprehensive cortical mapping, and (b) eliciting the response from the network of interest and not from adjacent cortical sites. Existing TMS targeting approaches frequently fail to precisely stimulate functionally relevant cortical areas, hindering treatment efficacy and the identification of biomarkers. The presented protocol integrates precise cortical mapping to acquire artifact-free TEPs, enabling reliable and reproducible measurements of early TEP components. This precision improves sensitivity to subtle neurophysiological variations and strengthens correlations with clinical phenotypes, supporting biomarker discovery in neuropsychiatric disorders. The proposed protocol utilizes structural, functional, and diffusion magnetic resonance imaging (MRI) to identify cortical patches belonging to the network of interest. Anatomical parcellation, functional connectivity, and real-time tractography are applied to locate areas with strong connectivity to other brain regions associated with the target network. The resulting personalized cortical clusters define the initial stimulation targets. TMS-EEG mapping is then employed to optimize TMS parameters by localizing cortical areas with high excitability, enhancing neuronal response magnitude while reducing non-neuronal noise, including muscle artifacts, decay, and other confounding factors affecting early TMS-EEG responses. A systematic exploration of the cortical mantle is conducted, adjusting stimulation location, orientation, and intensity, with continuous data quality monitoring through real-time visualization of averaged TEPs. TMS parameters producing artifact-free responses with clearly discernible early TEP components are selected for data collection. This article introduces the neuroimaging-guided TMS-EEG mapping technique and highlights the methodological advancements and benefits achievable through its application.
Primary Progressive Aphasias (PPA) are a group of neurodegenerative disorders characterized by the gradual decline of language abilities. They are typically divided into three major clinical variants: the non-fluent (nfvPPA), the semantic (svPPA) and the logopenic (lvPPA) variant. Even with an extensive clinical examination, a correct differential diagnosis among variants can be difficult due to the overlapping of dysfunctional language features. In this context, the combination of Transcranial Magnetic Stimulation and Electroencephalography (i.e., TMS-EEG) could extend our understanding of nfvPPA pathophysiology, given the possibility to non-invasively and directly measure cortical reactivity of brain speech networks to external perturbations. Twenty PPA patients (7 nfvPPA, 13 lvPPA) and 8 elderly controls underwent a TMS-EEG session targeting the left dorsal premotor cortex (Brodmann area 6). A subset of 9 patients (8 lvPPA, 1 nfvPPA) were additionally stimulated in the right homologous region. We automatically detected the EEG channel under the stimulator with the highest peak-to-peak amplitude of the early TMS-evoked response and computed the following measures: (i) natural frequency; (ii) normalized evoked spectral power in the alpha, low-beta, high-beta and gamma range. Non-fluent PPA patients showed a slower and simplified TMS-evoked response as compared to healthy elderly subjects, namely a reduction in high-beta power and natural frequency coupled with higher low frequencies (i.e., alpha) intrusion. No significant differences were detected between lvPPA and controls or nfvPPA and lvPPA. The speech rate was positively correlated with TMS-EEG measures (the high-beta power and the natural frequency). Furthermore, compared to the left side, the stimulation of the right hemisphere elicited TMS-evoked responses with higher natural frequency and high-beta power in both lvPPA and nfvPPA patients. This study first shows that TMS-EEG may provide useful neurophysiological biomarkers for characterizing the nfvPPA variant and monitor disease progression across variants. These findings might be employed in the future to stratify patients and eventually inform the application of variant-specific stimulation protocols tailored to individual neurophysiological profiles.
Translational network neuroscience aims to integrate advanced neuroimaging and data analysis techniques into clinical practice to better understand and treat neurological disorders. Despite the promise of technologies such as functional MRI and diffusion MRI combined with network analysis tools, the field faces several challenges that hinder its swift clinical translation. We have identified nine key roadblocks that impede this process: (a) theoretical and basic science foundations; (b) network construction, data interpretation, and validation; (c) MRI access, data variability, and protocol standardization; (d) data sharing; (e) computational resources and expertise; (f) interdisciplinary collaboration; (g) industry collaboration and commercialization; (h) operational efficiency, integration, and training; and (i) ethical and legal considerations. To address these challenges, we propose several possible solution strategies. By aligning scientific goals with clinical realities and establishing a sound ethical framework, translational network neuroscience can achieve meaningful advances in personalized medicine and ultimately improve patient care. We advocate for an interdisciplinary commitment to overcoming translational hurdles in network neuroscience and integrating advanced technologies into routine clinical practice.