Magnetic Resonance Imaging-Guided Temporal Interference Stimulation of the Cerebellar Fastigial Nucleus in Stroke Patients for Balance

Summary

This protocol describes a standardized, magnetic resonance imaging-guided temporal interference stimulation approach designed to selectively target the cerebellar fastigial nucleus in stroke patients. By enabling anatomically precise and reproducible deep cerebellar neuromodulation, this method seeks to enhance postural control and balance recovery.

Abstract

Lower limb motor and balance impairments are common dysfunctions after stroke. Although non-invasive brain stimulation has shown promise as an adjunct to neurorehabilitation, it remains limited to superficial cortical regions, instead of deep brain targets such as the cerebellar fastigial nucleus, which contributes to posture control and motor coordination. Temporal interference stimulation represents an emerging strategy for non-invasive brain stimulation for targeting deep neural structures. It delivers two high-frequency electric fields with slightly different carrier frequencies through the scalp. Computational models and early experimental studies suggest that temporal interference stimulation may preferentially modulate deep targets while reducing stimulation of superficial tissues. In this study, we introduce a standardized, image-guided protocol for cerebellar fastigial nucleus temporal interference stimulation in stroke patients. The protocol combines high-resolution structural magnetic resonance imaging, individualized electric field modelling, and computational optimization to design subject-specific electrode montages focusing toward the cerebellar fastigial nucleus. This protocol provides a reproducible framework for studying deep cerebellar neuromodulation and may support future mechanistic and rehabilitation studies of post-stroke motor and balance dysfunction.

Introduction

Lower limb motor and balance impairments are among the most common dysfunctions after stroke. These impairments remain major challenges for clinical rehabilitation. Despite advances in conventional rehabilitation strategies, recovery of postural control and gait is often incomplete. Non-invasive brain stimulation (NIBS) has therefore been investigated as a promising approach to stroke recovery. Clinical studies, including randomized controlled trials, suggest that techniques like repetitive transcranial magnetic stimulation (rTMS) can improve lower limb Fugl-Meyer scores and balance scales in chronic stroke patients by modulating neural networks involved in motor control^1,2,3.

The cerebellum is central to postural control and balance, with the fastigial nucleus (FN) serving as a critical deep output nucleus. The FN receives sensory input from the vestibular system and spinal cord, regulating trunk and proximal limb muscle activity through vestibulospinal and reticulospinal pathways⁴. Through these connections, the FN contributes to body stability during both static and dynamic states such as walking. Preclinical findings in animal models have shown that direct FN stimulation can promote neurological recovery, reduce ischemic brain injury, and support neuroplasticity^5,6,7,8. These findings support the FN as a biologically relevant target, but they do not establish efficacy in humans. However, this approach is invasive and would require surgery in humans, with potential risks including hemorrhage, infection, and hardware-related complications.

Traditional NIBS methods, such as transcranial direct current stimulation (tDCS) and rTMS, primarily modulate superficial cortical regions and provide limited spatial selectivity for deep cerebellar targes⁷. Temporal interference (TI) stimulation has been proposed as a non-invasive alternative for modulating these deeper structures⁹. Introduced by Grossman group, this approach applies multiple kilohertz currents through the scalp to generate an amplitude-modulated electric field in the brain^9,10. Critically, this modulation is not confined to deep regions but is distributed throughout the head, with its magnitude varying according to electrode configuration and individual anatomy^9,11. Computational studies and initial human studies suggest that the peak modulation can be shifted toward deeper structures more effectively than conventional transcranial stimulation^10,12. While early mouse models suggested direct neuronal activation in the hippocampus⁹. Subsequent studies have further supported the potential of TI stimulation to modulate human motor function^13,14,15 . The basic principle of this technique is illustrated in Figure 1. However, the precise underlying mechanisms, the achievable degree of selectivity, and the definitive clinical efficacy of temporal interference stimulation remain subjects of active investigation. Human studies have increasingly extended these findings to clinical settings. Specifically, TI stimulation has shown potential to improve memory by targeting the hippocampus in healthy older adults and patients with cognitive disorders¹⁰. In the motor dysfunction, it has also been reported to modulate the striatum and improve motor learning and coordination¹⁶. Although these findings are encouraging, the underlying mechanisms, the degree of selectivity, and the long-term clinical efficacy of TI stimulation remain unclear.

Although TI stimulation provides a modeling-based approach for deep targets, several key challenges remain for human application. In this protocol, individual anatomical variability is addressed by using high-resolution structural MRI to construct subject-specific head models. Electrode placement is optimized through computational field modeling to improve targeting of the cerebellar deep nucleus while limiting superficial exposure. Safety is addressed by standardized imaging-based stimulation, predefined stimulation parameters, and routine monitoring during stimulation. Thus, this approach provides a reproducible and individualized framework for deep cerebellar neuromodulation. It is intended to provide a foundation for future mechanistic research and targeted rehabilitation of post-stroke motor and balance dysfunction.

Protocol

Approval for the study protocol was obtained from the Institutional Review Board (IRB) or Ethics Committee. All study procedures involving human participants were conducted in strict compliance with the ethical principles outlined in the Declaration of Helsinki.

1. Participant Screening

1. Review the participant’s medical history to exclude contraindications, including pregnancy, epilepsy, implanted metallic or electronic devices (e.g., intracranial metallic foreign bodies, cardiac pacemakers, cochlear implants, etc.), intracranial hypertension, intracranial tumors, or skin lesions at electrode placement sites. In addition, exclude participants with severe cardiac, hepatic, or renal disease that may limit tolerance to the trial; those with psychiatric disorders such as major depression or schizophrenia; and individuals with skull defects shall be excluded^17,18,19.
2. Assess precautionary factors by administering a standardized non-invasive brain stimulation safety screening questionnaire prior to the session, including recent head injury, or the use of medications that may affect neural excitability (e.g., anticonvulsants)¹⁹.
3. Obtain written informed consent from all participants after providing comprehensive information about the study objectives, procedures, potential risks and benefits, alternatives, and participant rights, including the voluntary nature of participation and the right to withdraw at any time without any consequences.

2. MRI Data Acquisition and Processing for Targeting

1. Position the participant in a supine position within the MRI scanner, ensuring head stabilization using foam pads to minimize motion.
   CAUTION: MRI Risk. Ensure all metallic objects are removed from the participant prior to entering the MRI suite to prevent projectile hazards. Screen for claustrophobia and provide adequate hearing protection to mitigate scanner noise.
2. Acquire high-resolution T1-weighted structural images covering the entire head, including the scalp. The field of view should extend at least 10 mm beyond the head boundaries in the sagittal, coronal, and axial planes²⁰. Image resolution should be set to 1 mm isotropic, with a signal-to-noise ratio of no less than 20.
   NOTE: T1-weighted MRI is mandatory for all participants. Supplementary T2-weighted imaging should be obtained when T1-weighted MRI is insufficient for reliable visualization of posterior fossa tissue interfaces, cerebrospinal fluid spaces, or lesion/cavity boundaries. Supplementary CT should be obtained only when MRI-based skull segmentation is inadequate for accurate model construction.
3. Prepare the T1-weighted MRI data by ensuring the files are in the standard NIfTI format (.nii) and importing the data into a computational brain modulation platform through the designated upload interface. We used the brain modulation platform (as detailed in the Table of Materials) to perform whole-brain tissue modeling and automatically determine stimulation protocols from participants' MRI images.
4. Individualized Head and Brain Model Generation
   1. First, select the “Build Model” function from the modeling menu within the computational platform. Use the software's default preprocessing workflow to automatically correct for intensity inhomogeneity and perform spatial coordinate transformations on the MRI data²¹.
   2. Second, segment the participants' T1-weighted MRI images into six distinct tissue types (scalp, skull, cerebrospinal fluid, white matter, gray matter, and cavity) using the default tissue probability maps within the standard neuroimaging analysis software²².
   3. Visually inspect the segmented images slice-by-slice to verify the accuracy of the tissue boundary delineations. Next, assign the following conductivity values to each respective tissue type for the finite-element calculations: scalp (0.0002 S/m), skull (0.0202 S/m), cerebrospinal fluid (2 S/m), white matter (0.064 S/m), gray matter (0.103 S/m), and cavity (2.5e-14 S/m)²³.
   4. Third, the 10-10 EEG electrode system was registered onto the scalp models. Fourth, a tetrahedral finite element mesh was generated using iso2mesh²⁴.
   5. Finally, the finite element solver GetDP was employed to calculate the electric field intensity at each stimulation electrode under unit intensity, yielding the prior electric field matrix Leadfield²⁵. Expected output: a participant-specific 3D head model with segmented tissues, electrode positions, and a finite-element mesh.
5. Target Coordinate Definition and Spatial Registration
   1. Navigate to the coordinate input fields within the computational platform's targeting tool interface. Manually type the specific standard MNI space coordinates for the cerebellar fastigial nucleus (FN) into the respective X, Y, and Z text boxes (e.g., input [2, -55, -29] for the right FN and [-5, -55, -30] for the left FN).
   2. Execute the platform's automated spatial registration module. This algorithm first nonlinearly registers the participant's T1-weighted MRI image to the ICBM 152 Nonlinear atlases (2009) standard head model.
   3. Allow the software to automatically apply an inverse nonlinear spatial transformation. This step maps the pre-defined MNI target coordinates back into the participant's native T1 space, successfully generating personalized FN target coordinates within the individualized brain model for the subsequent electric field simulations.
6. Electrode Montage and Current Optimization
   1. Evaluate the final modeling result generated by the optimization algorithm. Accept the modeling result only if the software returns a valid electrode montage, the predicted FN envelope amplitude exceeds the predefined modeling reference level, and the hotspot is visually centered on or immediately adjacent to the individualized FN target.
   2. Classify the result as suboptimal (and adjust parameters to recalculate if necessary) if the FN intensity remains below this reference level, if the envelope appears spatially diffuse, or if the hotspot is displaced toward superficial cerebellar or occipital regions²⁶.
   3. Set the TI optimization parameters, including the predefined FN target coordinates, current constraints, and focality optimization mode. Accept the optimization result only if the software returns a valid montage with electrode positions, channel-specific current amplitudes, and a simulated field map showing maximal predicted envelope intensity at or near the FN target. In the present manuscript, a representative standard-head-model example is shown for workflow illustration (Figure 2).
   4. Evaluate the final modeling result generated by the optimization algorithm. Accept the modeling result only if the software returns a valid electrode montage, the predicted FN envelope amplitude exceeds the predefined modeling reference level, and the hotspot is visually centered on or immediately adjacent to the individualized FN target.
   5. Classify the result as suboptimal (and adjust parameters to recalculate if necessary) if the FN intensity remains below this reference level, if the envelope appears spatially diffuse, or if the hotspot is displaced toward superficial cerebellar or occipital regions.

3. Electrode Cap Preparation and Placement

1. Measure the participant’s head circumference from the brow ridge to the occipital protuberance using a flexible measuring tape.
2. Install circular rubber electrodes into the 72-position electrode cap at locations determined by the individualized electric field modeling results²⁷.
3. Seat the participant comfortably in an upright position with appropriate head support to minimize movement during stimulation.
4. Position the electrode cap according to the international 10–20 EEG system, ensuring accurate alignment with the nasion–inion line and bilateral preauricular landmarks. Adjust cap tension to achieve a secure fit without excessive compression²⁸.
5. Identify the targeted electrode sites through the empty electrode holders of the cap. Part the hair at each specific site using the wooden end of a cotton swab to fully expose the underlying scalp. Apply a small amount of abrasive paste to the cotton tip, insert it through the empty holder, and gently rub the exposed skin in a circular motion to remove superficial dead skin and reduce impedance.
   CAUTION: Scalp Preparation Risk. Apply abrasive paste carefully to avoid excessive skin abrasion, which can increase the risk of pain, skin burns, or infection during stimulation.
6. Inject conductive gel into each electrode until adequate contact with the scalp is achieved.
7. Participants are randomized to active TI stimulation or sham control groups using block randomization. Double-blinding is implemented: participants receive sham stimulation mimicking the active protocol, and experimenters conducting assessments are blinded to group allocation.

4. TI Stimulation System Setup and Parameter Adjustment

1. Connect the electrode cap to the TI stimulation device, which must support a minimum of two independent stimulation channels²⁹.
2. Open the device control interface and navigate to the stimulation parameter configuration menu. Manually type the carrier frequencies into their respective input fields: set Channel 1 to 2.000 kHz and Channel 2 to 2.005 kHz to yield a 5 Hz envelope frequency.
   1. Locate the current amplitude input fields within the interface. Use the interface controls to input the specific optimized current amplitudes for Channel 1 and Channel 2 exactly as generated by the computational platform.
   2. Visually cross-check the interface screen against the exported optimization report to verify all settings. Confirm that both channels appear as "active" and all input parameters are correctly listed on the main control panel before proceeding.
3. Activate impedance monitoring on the stimulation device and adjust electrode contacts until all channel impedances fall below 8 kΩ³⁰. If impedance remains above 8 kΩ, repeat scalp preparation and adjust electrode contact before continuing.
4. Deliver a brief pre-stimulation phase to assess participant tolerance.
   CAUTION: Electrical Stimulation Risk. Always initiate stimulation at a low intensity and ramp up gradually to prevent sudden shock or distress.
5. If necessary, proportionally adjust the current amplitude while maintaining the relative balance between channels. Proceed to the formal session only if the participant reports no intolerable discomfort. If moderate or severe discomfort occurs, reduce current proportionally across channels or stop stimulation. The stimulation settings for the active and sham conditions are summarized in Table 1.

5. Stimulation Procedure

1. Prior to initiating the stimulation, instruct the participant to remain seated quietly, minimize head and body movements. After confirming all stimulation parameters, initiate the formal stimulation session by selecting “Start” on the device interface. Sinusoidal waveforms are delivered through both channels to generate temporal interference at the cerebellar FN.
2. Continuously monitor electrode impedances via the device interface to ensure they remain within safe operational limits. Prior to initiating the session, instruct the participant to spontaneously report any adverse effects or discomfort, such as unexpected tingling sensations or headache, at any time during the stimulation.
   CAUTION: Electrical Stimulation Risk. Electrical stimulation carries a risk of scalp burns or excessive pain.
3. Immediately pause or terminate stimulation if electrode impedance exceeds 8 kΩ or if the participant reports severe discomfort³¹.
4. Upon completion of the stimulation session, terminate stimulation by selecting “Stop,” remove the electrode cap carefully, cleanse the scalp with water, and inspect the skin for signs of irritation.

Results

Electric Field Modeling Example
To evaluate whether the individualized MRI-guided workflow could achieve the study objective of steering temporal interference stimulation toward the cerebellar fastigial nucleus (FN), representative results were analyzed in a logical sequence from model generation to field localization. High-resolution T1-weighted MRI data were successfully segmented into six tissue compartments, enabling individualized electric field modeling for subsequent montage optimization³².

The main modeling outcome was that computational optimization generated a valid two-channel electrode montage for FN targeting (Figure 3A). Under this montage, although the high-frequency carrier fields remained strongest in superficial regions, the low-frequency amplitude-modulated (AM) envelope was centered on or immediately adjacent to the FN target (Figure 3B). The predicted AM field intensity within the FN exceeded 0.2 V/m, indicating that the workflow was able to generate a focal deep-target solution rather than a predominantly superficial distribution^9,10. In the present study, this value was used as a pragmatic modeling reference rather than a validated physiological threshold for human FN modulation. Prior optimization studies have also used 0.2 V/m as a minimal electric-field constraint, while noting that the effective modulation threshold of temporal interference stimulation remains uncertain and may vary across targets and individuals^32,33. Together, these findings support the feasibility of using individualized MRI-guided modeling to steer TI stimulation toward the cerebellar FN.

Examples of suboptimal outcomes and troubleshooting
In practice, not all individualized TI stimulation sessions produce an optimal focal envelope at the cerebellar FN on the first attempt. A suboptimal modeling result may present as a diffuse AM envelope, insufficient field intensity at the FN, or displacement of the hotspot toward superficial cerebellar or occipital regions. Such findings may be related to imperfect tissue segmentation, registration inaccuracy between template and individual anatomical space, limited candidate electrode configurations, or inter-individual anatomical variability^13,27. In these cases, the operator should re-evaluate MRI quality, tissue segmentation, and target registration, and repeat montage optimization using alternative electrode pairs if needed.

During stimulation setup, another common suboptimal condition is persistently elevated electrode impedance or unstable channel contact. This is typically associated with poor electrode–skin contact, inadequate conductive medium, or hair interference, and should be addressed by re-cleaning the scalp, adding conductive gel, adjusting electrode position and pressure, and confirming cable integrity before proceeding^34,35.

Mild transient discomfort, such as tingling, itching, burning sensation, or scalp irritation, may also occur during pre-stimulation or stimulation onset. When these effects remain tolerable, stimulation parameters may be adjusted and the participant should be closely monitored; stimulation should be paused or discontinued if discomfort persists or safety concerns arise^36,37,38.

Figure 1. Basic principle of TI stimulation
Temporal Interference (TI) stimulation utilizes two pairs of scalp electrodes to deliver high-frequency currents (I₁ and I₂) with a marginal frequency offset, establishing oscillating electric fields (E₁ ​and E₂​) within the brain. The superposition of these fields in specific deep-seated regions creates a low-frequency AM envelope. This technique facilitates the selective activation of neurons at depth while bypassing superficial cortical areas, which remain unresponsive to the high-frequency components. Please click here to view a larger version of this figure.

Figure 2. Workflow for participant-specific electric field simulation. (A) Acquisition of T1-weighted structural MRI images. (B) Tissue segmentation. (C) Registration of 10-10 EEG electrode position. (D) Generation of a tetrahedral finite-element mesh. (E) Computation of the resulting intracranial electric field distribution. Please click here to view a larger version of this figure.

Figure 3. Electric field intensity in a representative standard-head-model simulation.
(A) Schematic illustration of the optimized electrode montage. Two pairs of electrodes delivering high-frequency carrier currents (2 kHz and 2.005 kHz) are positioned on the occipital scalp and driven by the NervioX-1000 stimulator. (B) Electric field intensity for a typical participant. EMAX: maximum electric field intensity in target region, corresponding to the red cross line. The modeling process included brain tissue segmentation, electrode placement (10-10 EEG system), finite element meshing, and TI electric field solution. A = Anterior; P = Posterior; R = Right; S = Superior. Please click here to view a larger version of this figure.

Figure 4. Representative findings from previous TI studies. (A) Focal modulation of hippocampal blood-oxygen-level-dependent (BOLD) responses and memory performance in humans using TI stimulation. (B) Frequency-specific modulation of endogenous gamma oscillations in the rat hippocampus, demonstrating selective entrainment without activation of overlying cortical areas. (C) Frequency-dependent enhancement of human memory encoding, showing that only the physiologically relevant difference frequency (Target Freq) produces significant behavioral improvement compared with sham stimulation and unrelated frequencies. (D) Non-invasive modulation of the human striatum. Theta-burst patterned tTIS increased BOLD activity in the putamen and enhanced motor skill acquisition compared to high-frequency controls. (E) Direct measurement and validation in primates. Intracranial recordings in rhesus monkeys confirmed that TI generates sufficient envelope modulation amplitudes in deep brain regions. Pilot data also showed reduced resting tremor in Parkinson's disease patients following TI. (F) Enhancement of lower limb motor performance. Repetitive TI stimulation targeting the primary motor cortex leg area significantly improved vertical jump height (countermovement jump and squat jump) in healthy adults. This schematic was created by the authors based on published studies and does not reproduce original published figures. (A) adapted from Violante, I.R. et al (2023)¹³, (B) adapted from Esmaeilpour, Z. et al. (2021)⁴⁰, (C) adapted from Missey, F. et al (2026)⁴¹, (D) adapted from Wessel. et al. (2023)⁴², (E) adapted from Liu, R. et al. (2024)⁴³, (F) adapted from Zheng, S. et al. (2024)¹⁵. Please click here to view a larger version of this figure.

Figure 5. Simulation of the TI stimulation protocol for the Cerebellar Fastigial Nucleus (FN). 
Targeted simulation of the FN [Right (2, -55, -29); Left (-5, -55, -30)] was performed using the NervioWeb platform (China, version 1.1.0). The results demonstrate that an optimized combination of two electrode pairs (P9/O9 at 2.0 mA; FT10/O10 at 1.292 mA) can effectively generate the desired electric field envelope at the FN target. Note: This simulation protocol is for standard head models. Individualized modeling is recommended for specific patients. A = Anterior; P = Posterior; R = Right; S = Superior. Please click here to view a larger version of this figure.

Condition	Electrode montage	Carrier frequencies	Envelope frequency	Current amplitude	Total stimulation duration	Sham delivery mode	Blinding assessment
**	**	**	**	**	**	**	**
**	**	**	**	**	**	**	**

Table 1. Stimulation parameters for active and sham temporal interference stimulation.

Discussion

Methodological overview and key implementation considerations
This study presents a standardized protocol for the non-invasive modulation of the cerebellar FN using individualized TI stimulation. By integrating high-resolution structural MRI, computational electric field modeling, and a multi-channel stimulation interface, the proposed protocol enables anatomically informed and precise targeting of the FN. This approach establishes a novel methodological framework for investigating the functional contributions of deep cerebellar nuclei to motor control and balance without the risks inherent to invasive neurosurgical procedures. Unlike general descriptions of TI that focus on broad principles, our protocol emphasizes the specific anatomical constraints of the posterior fossa and the necessity of individualized optimization for reaching the FN.

A critical determinant of successful FN targeting in this protocol is the individualized electric field modeling and subsequent optimization of the electrode montage. The FN is a small, deeply situated structure embedded within a complex anatomical environment characterized by variations in cerebrospinal fluid distribution and skull thickness, rendering generic electrode configurations insufficient for reliable stimulation¹⁶. High-quality T1-weighted MRI data, and optionally T2-weighted images, with at least 1 mm isotropic resolution are therefore essential to ensure accurate tissue segmentation and model fidelity. As detailed in the protocol, finite element method solvers are required to simulate the spatial distribution of the interference envelope³⁹, and investigators must verify that the maximal envelope amplitude is localized to the FN rather than to superficial cortical regions.

To address practical challenges during implementation, we recommend specific troubleshooting strategies. If simulation results indicate off-target stimulation, investigators should iteratively adjust electrode positions towards the orthogonal plane of the target axis. In cases of poor field convergence, refining the finite element mesh density in the region of interest can improve model accuracy. Furthermore, careful scalp preparation during experimental setup remains essential. Although the use of high-frequency carrier currents reduces cutaneous sensation, elevated electrode impedance may still result in tingling or pruritic sensations. Thorough scalp abrasion using conductive preparation agents (e.g., Nuprep) to maintain impedance below 8 kΩ is necessary. To further preserve blinding integrity, we suggest employing active sham protocols or applying topical anesthetics (e.g., EMLA cream) to electrode sites to mask somatosensory differences between active and sham conditions.

Supporting evidence from prior TI studies
Although FN-specific behavioral and physiological outcome data from the present protocol are forthcoming, the capability of TI stimulation to engage deep neural targets has been consistently supported by recent experimental and computational studies (Figure 4). In a seminal human study, Violante et al. demonstrated that TI stimulation targeting the hippocampus significantly modulated task-evoked blood-oxygen-level-dependent (BOLD) responses and enhanced episodic memory performance, providing direct evidence for effective deep brain neuromodulation without concurrent activation of overlying cortex (Figure 4A)¹³. Complementary rodent studies have shown that TI stimulation can selectively modulate endogenous hippocampal gamma oscillations while sparing superficial cortical regions, thereby validating the “pass-through” property of high-frequency carrier fields (Figure 4B)⁴⁰. Frequency-dependent effects have been observed; for instance, Missey et al. reported that TI stimulation promotes memory encoding only when the difference frequency matches the physiological rhythm of the target network (Figure 4C)⁴¹. To further substantiate the feasibility of targeting deep cerebellar structures, we performed a computational simulation using the computational brain modulation platform. Our results demonstrate that optimized TI stimulation can effectively steer the electric field envelope to the cerebellar FN (Figure 5). To provide definitive physiological validation, we are concurrently acquiring SEEG in epilepsy patients with cerebellar electrodes, these recordings are intended to provide electrophysiological measurements associated with the stimulation field.

Extending these findings to motor systems, which is highly relevant for cerebellar applications, recent investigations have provided further validation of the efficacy of TI stimulation in modulating deep motor-related structures (Figure 4D–F). Wessel et al. applied theta-burst patterned TI stimulation to the human striatum and demonstrated, using functional MRI, enhanced activation within the striatum and associated motor networks, accompanied by significant improvements in motor skill learning, without co-activation of superficial cortical areas (Figure 4D)⁴². In non-human primates, Liu et al. directly quantified the spatiotemporal properties of TI induced electric fields using stereo-electroencephalography (SEEG), confirming robust envelope modulation amplitudes in deep brain regions. Preliminary clinical observations further suggested that targeted TI stimulation may alleviate resting tremor in patients with Parkinson’s disease (Figure 4E)⁴³. In addition, a randomized controlled study by Zheng et al. demonstrated that repetitive TI stimulation applied to the lower limb motor cortex significantly increased vertical jump height in healthy male participants, highlighting the potential of TI stimulation to modulate motor performance (Figure 4F)¹⁵.

Taken together, evidence spanning biophysical modeling, neurophysiological measurements, neuroimaging, and behavioral outcomes across cognitive and motor domains provides converging support for the feasibility of focal, frequency-specific neuromodulation using TI stimulation. These findings collectively support the premise that the present protocol can effectively target and modulate neural activity within the human cerebellar FN.

Comparative advantages of FN-targeted TI stimulation
Relative to existing non-invasive stimulation approaches, our computational modeling results suggest that the present TI protocol offers distinct advantages in both penetration depth and spatial steerability⁴³. Conventional tES methods, including tDCS and transcranial alternating current stimulation (tACS), are limited by substantial current shunting through the scalp and skull, which restricts effective modulation primarily to superficial cortical targets. Although Deep Brain Stimulation (DBS) remains the gold standard for focal stimulation of subcortical nuclei, its invasive nature limits widespread application ⁴⁴. The TI approach described here partially bridges this methodological gap. Theoretical simulations indicate a depth-to-surface efficacy ratio that exceeds that of conventional tES. Importantly, this protocol allows electronic steering of the stimulation focus by adjusting the relative current amplitudes between electrode pairs, for example by modifying the current ratio from 1:1 to 1:1.5, without physically repositioning electrodes⁴⁵. This capability provides a practical means of compensating for individual anatomical asymmetries, a level of flexibility that is not achievable with standard tES montages ⁴⁵.

Limitations of the approach
Despite these advantages, several limitations of the present approach warrant consideration. First, although TI offers improved depth targeting compared with tDCS, its spatial resolution (quantified by the full width at half maximum of the envelope amplitude) remains on the order of centimeters and does not approach the millimeter-scale precision afforded by implanted DBS electrodes. Recent advancements such as multipolar TI (mTI) have demonstrated the potential to further sharpen focality⁴⁶. While mTI may offer superior selectivity for extremely small nuclei, we posit that the standard TI protocol described here provides a balanced trade-off between complexity and focality for the FN, given its anatomical dimensions. As a result, unintended modulation of adjacent cerebellar nuclei or nearby white matter pathways cannot be entirely excluded. Second, the physiological effects of TI stimulation are critically dependent on the accuracy of the underlying computational model. Uncertainties in tissue conductivity parameters, particularly those related to skull and cerebrospinal fluid properties, may lead to discrepancies between simulated and actual electric field distributions. In the present study, we did not perform formal uncertainty quantification or sensitivity analysis across alternative conductivity values. Third, although high-frequency carrier currents substantially reduce peripheral sensation, they do not eliminate it in all individuals, which may complicate the implementation of fully double-blind, sham-controlled experimental designs.

Future perspectives and translational potential
Future applications of this protocol may extend beyond the investigation of motor control and balance. The ability to non-invasively target deep cerebellar nuclei holds promise for therapeutic exploration in disorders characterized by cerebellar dysfunction, including ataxia, post-stroke motor impairment, and potentially cerebellar-related cognitive and affective syndromes⁴⁵. Future work may further improve the present MRI-guided TI pipeline by incorporating patient-specific whole-brain of virtual brain twin models that integrate structural imaging with electrophysiological data, thereby refining target definition and stimulation parameter optimization at the individual level⁴⁷ . Methodological developments should prioritize integration with concurrent neuroimaging techniques, such as functional MRI or electroencephalography, to empirically verify target engagement and establish closed-loop relationships between stimulation parameters and physiological responses. Moreover, further refinement of computational models, particularly through the incorporation of anisotropic conductivity properties of cerebellar white matter tracts, is likely to enhance targeting precision for deep nuclei such as the FN.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
72-channel Electrode Cap	Suzhou NeuroDome Medical Technology Co., Ltd	Suzhou, China	72-point electrode cap for brain stimulation
Conductive Gel	Not applicable	Not applicable	Gel for enhancing electrode-skin contact
Flexible measuring tape	Not applicable	Not applicable	Used to measure head circumference for electrode cap fitting
NervioWeb software platform	Suzhou NeuroDome Medical Technology Co., Ltd	Suzhou, China	Online brain modulation experimental platform
NervioX-1000 stimulator	Suzhou NeuroDome Medical Technology Co., Ltd	Suzhou, China	Non-invasive deep brain stimulation system
Nuprep skin prep gel	Weaver and Company	Aurora, CO, USA	Abrasive paste for skin cleaning to enhance conductivity