A Methodological Protocol and Considerations for Transcranial Ultrasonic Stimulation in Exploratory Clinical Human Studies

A unique and desirable combination, the deep, focal, and non-invasive nature of transcranial ultrasonic stimulation¹ (TUS) has spurred growing interests in exploring its therapeutic potentials for neurological and neuropsychiatric diseases (e.g., stroke², Parkinson's disease³, depression⁴, Alzheimer's disease⁵, etc.). Expert consensus and guidelines have sought to improve the reproducibility and interpretability of TUS research^6,7. However, these primarily address the broader basic research applications of TUS and offer less guidance on exploring therapeutic relevance in humans. For example, when comparing basic and clinical TUS research, the total stimulation duration (TSD) differs substantially - often up to around 80 s in basic research^8,9 versus over 10 min in exploratory protocols designed to approximate clinical use². Such distinctions underscore the need for tailored guidance when exploratorily applying TUS to a neurologically diseased cohort. While this protocol focuses on TUS in stroke, its framework is broadly applicable to other patient populations, as well as neurotypical adults.

TUS delivers pressure waves to the target brain tissue for modulation⁷. Essential TUS hardware components include function generator(s), an amplifier, a transducer, and, especially for extended TSDs, an on-head transducer mount⁶. The function generator produces the electrical waveform that defines the exiting pressure pattern emitted by the transducer. While some function generators can generate the peak voltages required for TUS transducers, most require an amplifier to boost the signal between the function generator(s) and the transducer. The amplifier increases the amplitude of the electrical waveform to the level needed for effective ultrasound generation. Commercially available TUS systems often integrate the function generator and amplifier into a single turnkey unit, allowing direct connection to the transducer. The transducer, typically made of piezoelectric materials, converts the electrical signals into mechanical vibrations to produce ultrasound waves. Although manual transducer holding is possible, this approach is labor-intensive and susceptible to displacement errors, while a minor displacement can cause complete mistargeting of the intended tissue⁷. Mechanical arm with chin rest is also used and works well for basic research with relatively shorter TSDs. A feasible alternative for exploratory clinical TUS with TSDs over 10 min is 3D printing a transducer mount equipped with flanges for secure strap fixation (Figure 1), which can balance positional stability with ease of use in exploratory clinical settings and participant comfort.

Figure 1: 3D-printed mount for TUS transducer. Single-element transcranial ultrasonic stimulation (TUS) transducer (center piece pointed by the red arrow) housed in a 3D printed transducer mount (pointed by blue the arrow) with top openings (pointed by the green arrow) for neuronavigation's tool tracker, side openings (pointed by the brown arrow) to secure TUS transducer and tool tracker(s) using screws, and flanges for a strap-based on-head fixation. Please click here to view a larger version of this figure.

Optional hardware components include an electrical impedance matching circuit, a power meter, an oscilloscope, a neuronavigation system, and a thermocouple. These devices respectively aid in reducing electrical reflections to optimize power transfer, indirectly verifying acoustic outputs (power meter and oscilloscope), ensuring spatial accuracy, and monitoring safety. Their implementation details and methodological considerations are elaborated in the Discussion and the Protocol sections.

Optimizing the parameters of transcranial ultrasound stimulation is an active and evolving area of research, which pertains to the design of the electrical drive waveform. The vertical axis of a drive waveform determines the amplitude of transducer's pressure output. The horizontal axis of a drive waveform determines the timing of the transducer's pressure output, which can be broken down into a few duration and repetition types. Using the nomenclature outlined in the reporting guidelines by the International Transcranial Ultrasonic Stimulation Safety and Standards Consortium (ITRUSST)¹⁰, TUS timing parameters are organized from the shortest to the longest durations as follows: ramp duration (RD) is the time during which the pressure output ramps up from zero to the prescribed maximum; pulse duration (PD) is the duration for a single, active, and continuous pressure output; pulse repeat interval (PRI) is the duration between the beginning of two adjacent PDs, consisting of one PD and one "off time" during which the transducer output is zero, and the reciprocal of PRI is the pulse repetition frequency (PRF); pulse train duration (PTD) is the duration for which the PRI is repeated; pulse train interval (PTI) is the duration between the beginning of two adjacent PTDs, consisting of one PTD and one inter-PTD interval during which the transducer output is zero; pulse train repeat duration (PTRD) is the duration for which the PTI is repeated.

Figure 2: TUS waveform illustration. Typical transcranial ultrasonic stimulation waveforms for potential clinical efficacies. The nomenclature adheres to the ITRUSST reporting guidelines¹⁰. RD, ramp duration; PD, pulse duration; PRI, pulse repeat interval; PRF, pulse repetition frequency; PTD, pulse train duration; PTI, pulse train interval; PTRD, pulse train repeat duration. Please click here to view a larger version of this figure.

Figure 2 illustrates these timing parameters. Table 1.1 reports the timing parameters in Huang et al., 2025², from which this protocol is derived, in accordance with the tabular form recommended by the ITRUSST reporting guidelines¹⁰ while Table 1.2 reports additional parameters. Auditory effects are widely recognized as a major confound in TUS experiments^6,11. Ramping the drive waveform is the most used method to mitigate auditory effects and thus will be considered the primary mitigation method in this protocol. Theoretically, distinct ramping profiles can be applied at the start and end of each active pressure output (i.e., PD, PTD, PTRD), but most studies employ a same symmetrical PD repeated throughout the entire stimulation session, resulting a single RD in the drive signal. Ramping is implemented using window functions, which define the "shape" and "speed" of the signal going from zero to the prescribed maximum, with the Tukey window being the most commonly used in the literature¹⁰. After controlling for auditory confounds, selecting parameters to elicit excitatory versus inhibitory effect remains an area of active research and depends on the specific experimental question. Parameter choices in Huang et al., 2025², from which this protocol is derived, were informed by existing literature¹², with a PD decreased from 0.36 ms to 0.2 ms for additional safety margins. Methodological aspects were adapted from Legon et al., 2018¹³. These selections are one detailed example rather than a prescriptive standard, reflecting the still-exploratory stage of clinical TUS. Consult the relevant literature for guidance tailored to study-specific hypothesis and objectives.

Transducer calibration is a critical consideration of TUS, and water tank scanning represents the most comprehensive method currently available. For non-turnkey TUS systems, an initial water tank scanning is strongly recommended to establish the transmitting sensitivity of the transducer, i.e., the relationship between the electrical drive voltage in Volts and the resulting pressure output in Pascals. Turnkey systems often come with vendor-preconfigured settings that may reduce the need for an initial calibration, but verifying the transducer output ensures accurate and high-quality TUS delivery. Similarly, although subsequent calibrations can be minimized if necessary, regular checks are advisable to enhance study rigor.

Performing water tank scanning generally requires specialist equipment and training. This protocol provides a brief overview in Step 2 (based on Chen et al., 2023¹⁴). Alternatively, the procedure can be outsourced to a collaborative laboratory with the appropriate expertise with confirmation that all required parameters are collected (Step 2.7). Power meter measurements provide a simple, indirect method for measuring transducer output. After calibration or vendor characterization, power meter readings can serve as a benchmark for pre-administration output. Variation in power output implies corresponding changes in pressure outputs, indicating a need for either re-calibration or adjustment of the drive signal to maintain consistent stimulation.

TUS targeting involves determining the optimal transducer placement, both location and orientation, to effectively stimulate the tissue of interest. The current state-of-the-art approach employs numerical simulations to identify the optimal transducer placement (Step 3 of the protocol), though this represents one example rather than a definitive standard. Alternatively, an empirically based method previously reported locates the scalp location where transcranial magnetic stimulation (TMS) most effectively elicits contralateral motor responses, and positions the transducer at that location with its surface normal to the scalp^2,3 (Step 4). In principle, this TMS-guided approach can be extended to other brain areas that have an identifiable output, such as the sight of phosphenes for the visual cortex. However, more data/publications are needed to validate the broader applicability of such approach.

Outcome measures for clinical TUS include both safety and efficacy-based assessments. Given the exploratory stage of clinical translation, pronounced efficacy may not yet be detectable using standard clinical measures. Therefore, surrogate response indicators are often used in place of efficacy measures to detect effects below clinical efficacy. Safety monitoring can incorporate both online and offline assessments. Online safety can be tracked using thermocouple(s) placed on the scalp during stimulation, offering a practical approach in the clinical setting (Step 6.5). For offline safety, this protocol provides instructions in apparent diffusion coefficient (ADC) quantification (Step 8), as ADC is commonly used to evaluate tissue integrity¹⁵. Surrogate response indicators vary depending on the targeted disease or domain of interest. For studies of post-stroke motor recovery, contralateral corticospinal excitability (CSE) and motor sequence learning (MSL) measures will be described (Steps 9-11) as they provide reliable neurophysiological and functional markers for the exploratory clinical efficacy of TUS.

Taken together, this protocol provides a detailed, modular, and adaptable example for administering TUS in both neurological and neurotypical adults, emphasizing ease of use and extended stimulation duration. The protocol is derived from Huang et al., 2025², which employs a non-turnkey drive system that allows customization of pulse shaping and single-element transducer. In a step-by-step manner, the system-agnostic general principle is presented first, and, where appropriate, system-specific example implementation is given.

Experimental procedures presented in Huang et al., 2025² have been approved by the Duke Institutional Review Board and are in accordance with the recently updated Declaration of Helsinki. Obtain informed consent from each participant before enrollment.

NOTE: Follow the instructions while simultaneously working with the hardware/software to enhance understanding and ensure accurate implementation. Figure 3A shows an overview of the steps involved, Figure 3B shows a decision flowchart for choosing simulation-based (Step 3) versus non-simulation-based (Step 4) targeting. Supplementary materials provide sample pre- and post-TUS checklists covering eligibility for consent, safety screening, TUS equipment quality control, and adverse event monitoring.

1. Consenting and safety screening

1. Inclusion criteria: i) ≥ 21 years old of any race or sex, ii) first-ever ischemic or hemorrhagic stroke over 1 month ago, iii) presented with unilateral limb weakness and had a Fugl-Meyer Upper Extremity scale ≤ 62 (maximum score = 66), and iv) paretic abductor pollicis brevis (APB) has motor evoked potentials (MEPs) that can be elicited by TMS. Exclusion criteria: i) arm functions affected by any concomitant neurological disorder(s), ii) documented severe dementia prior to stroke onset, regardless of medication status, or iii) any contraindication(s) to magnetic resonance imaging (MRI)¹⁶/TMS¹⁷/TUS¹⁰.
2. Inform each participant of the purpose(s) of the study, the primary procedures involved, and any associated risks. Respond to any inquiries or concerns they may express, then have them confirm their understanding of the consenting process by signing the consent form.
3. Survey each participant with the TUS¹⁸ and TMS¹⁷ safety checklist. Exclude participants who do not fulfill all safety requirements.

2. TUS transducer characterization ¹⁴

NOTE: Perform the steps in this section to obtain an initial characterization of a new TUS transducer, or, at regular intervals, a confirmation/calibration for the input voltage-output acoustic pressure relationship of the transducer. Figure 4 shows the experiment setup for this water tank-based characterization of a TUS transducer.

1. Fill a water tank with degassed, deionized water to prevent potential damage to the hydrophone.
2. Fixate the ultrasound transducer onto the water tank's fixture and mount the hydrophone onto the 3-axis positioning system.
3. Connect the hydrophone to an oscilloscope, and switch on the power of the hydrophone only after it is fully immersed in water.
4. Drive the transducer using a function generator and radio frequency power amplifier, only after the transducer is fully immersed in water and the hydrophone is powered on.
   NOTE: Ensure the input electrical waveform contains at least 5 cycles for each pulse (i.e., PRI and PTI) to stabilize the transducer output.
5. Position the hydrophone tip at the visually defined approximate focal region of the ultrasound transducer using a 3-axis positioning system, scan the hydrophone across the water tank at locations of possible primary pressure output, record oscilloscope reading at each location to obtain the acoustic pressure distribution, and record the hydrophone location at which the focal region is found.
   NOTE: Calculate acoustic pressures using the hydrophone's sensitivity (i.e., mV/MPa) and the oscilloscope reading of the hydrophone-recorded electrical signal.
6. Assess the transmitting sensitivity of the ultrasound transducer (i.e., pressure output of the transducer over peak-to-peak amplitude of input voltage, kPa/V_PP) at the focal point by applying different input voltages.
7. Confirm that the transmitting sensitivity (kPa/V_PP) of the transducer, the focal depth of the transducer (mm), and the -6 dB beamwidth of the ultrasound beam (mm) are all collected if outsourcing transducer characterization to a collaborative laboratory with the appropriate expertise. The focal depth defines the maximum penetration depth at which the potentially therapeutic energy can be effectively delivered to the targeted brain region. The -6 dB beamwidth of the ultrasound field specifies the lateral extent of the focal region, thereby delineating the effective treatment volume within the brain tissue.

3. Simulation-based TUS targeting (recommended, participant scan required)

NOTE: Simulation-based TUS targeting requires a predetermined intracranial location to be stimulated. Such target-selection is study specific, and commonly used methods include anatomical approaches based on MRI atlases and functional approaches guided by individual functional MRI activation data⁷, as well as connectivity-informed targeting¹⁹. Select study- and hypothesis-appropriate method and follow the relevant literature^7,19 to determine the intracranial target. Next, follow the steps below to obtain the TUS transducer placement for stimulating the intracranial target.

1. Reconstruct a segmented head model from the MRI - example implementation: Install the electric field (e-field) simulation software (see Table of Materials) if not already installed, and open a terminal at the folder of participant's structural MRI scan. Run the CHARM pipeline of the e-field software²⁰ by typing charm ParticipantID ./ParticipantID_T1.nii.gz and hitting Enter/return on the keyboard. If a T2 is available, charm ParticipantID ./ParticipantID_T1.nii.gz ./ParticipantID_T2.nii.gz can be run instead. Note the spaces and periods in the commands above. Depending on the kind of terminal, the forward slash (/) may need to be replaced with backward slash ().
2. Check segmentation results, especially for the lesion (if any) and the skull - example implementation: Check CHARM's segmentation results by opening the charm_report.html in the m2m_ParticipantID folder created, and clicking on the viewer to move through slices. Confirm especially the outline for cortical and trabecular bones, the brain (grey and white matters), and the scalp, as the downstream TUS targeting software²¹ (see Table of Materials) uses primarily these. Follow the e-field software's guide to manually edit the head model if the segmentation is unsatisfactory.
   NOTE: For participants with a lesion such as stroke, the e-field software allows custom delineation of a lesion and thus custom segmentation assignment. However, the TUS targeting software, as of V0.4.3, has not yet been able to recognize custom segments from the e-field software. Therefore, when a lesion is present, it suffices to only confirm that the outlines for the aforementioned tissue types are correct and not affected by the lesion.
3. Setup for intracranial target definition - example implementation: Hit New SimNIBS Project… in the neuronavigation software (see Table of Materials) and select the .msh file in the m2m folder.
   NOTE: Version 2.5.9 of the neuronavigation software was used to detail this and the following steps, but these steps are applicable to versions from 2.5.9 to 2.5.3.
4. Setup for physical head-MRI registration - example implementation: Hit Configure Landmarks… under the Landmarks tab, place landmarks on the tip of the nose, the nasion, and left/right supratragic notch, then save the project file.
   NOTE: Landmarks for this step should be where the skin is not compressed or stretched during the MRI. Choose landmarks other than the ones above as desired, depending on the MRI acquisition of their participants.
5. Define an intracranial target and load it to a TUS targeting software - example implementation: Hit Configure Targets… under the Targets tab and bring Scalp & Targets, Inline & Targets, and Inline 90 & Targets to view. Left click on the intended TUS focus inside the brain, project an offset cursor to the scalp using the Crosshairs Offset slider at the lower right, slide the AP slider on the right to minimize the scalp to target distance in Inline & Targets, then slide the Lat slider until the virtual transducer is tangential to the scalp in Inline 90 & Targets. At the New… dropdown menu (upper left of the window), select Trajectory to save this placement. At the lower left of the window, toggle fUS, select a results folder, then hit Compute Simulation to invoke the TUS targeting software.
   NOTE: Check carefully if there is any anatomical sinus/air-filled cavity along the path of the trajectory created in this step. If so, choose the next best trajectory avoiding the sinus/cavity.
6. Initialize the TUS targeting software - example implementation: Confirm the auto-populated entries in the TUS targeting software under Imaging input are correct except for the thermal profile, select the transducer that will be used ("Single" means single-element transducer) and, if more than one computing backends are available, select the desired computing backend. Next, hit Select Thermal Profile … to load the correct .yaml file that describes the temporal pattern of TUS driving signal. Refer to the TUS targeting software's documentation for the construction of this .yaml file. As an example, the .yaml file for the TUS temporal pattern in Huang et al., 2025² would be
   BaseIsppa: 8.0 # W/cm²
   AllDC_PRF_Duration: #All combinations of timing that will be considered
    - DC: 0.2
    PRF: 1000.0 #Hz
    Duration: 0.5 #s
    DurationOff: 1.0 #s
    Repetitions: 480
   NOTE: This example is for illustrative purposes only. 480 repetitions would be too excessive for the TUS targeting software to compute. This .yaml file affects only the thermal simulation of the TUS targeting software and does not affect its acoustic simulation, while the acoustic simulation is the only component that TUS targeting is based on. For long TUS, use the TUS targeting software for targeting and not thermal simulation. Use the open-source acoustic simulation software (see Table of Materials) for thermal simulations of long-duration paradigms. Refer to its documentation to estimate temperature rise, cumulative equivalent minutes at 43 °C (CEM43) thermal dose, and thermal index, ensuring the selected parameters comply with Food and Drug Administration²² and ITRUSST biophysical safety¹⁸ guidelines.
7. Obtain tissue masks and input parameters for acoustic simulation - example implementation: Hit CONTINUE to get to Step 1 - Calculate Mask and select the correct US Frequency (kHz) (i.e., the transducer's fundamental frequency) and the desired PPW (i.e., points per wavelength). 9 PPW ensures stable and accurate results but incurs a heavier compute burden, and the default 6 PPW is the TUS targeting software's minimum that balances fast computation, stability, and accuracy. Hit Calculate planning mask. After the computation finishes, inspect the output mask to ensure the scalp, bone, and brain are labelled to desirable accuracy.
8. Input transducer characterizations and run the TUS targeting - example implementation: Hit Step 2 -Ac Sim. If Single was selected at Step 3.6, input Focal length (mm) and Diameter (mm) of the transducer, input the distance between scalp and the transducer's exit plane (due to coupling media such as spread gel or gel pad) at Distance Tx outplane to skin (mm), and input Max. depth beyond target (mm). This depth is the space beyond the intended target to include in the simulation. The default 40 mm suffices in most cases, but if the intended target sits above bony tissue, increasing this depth ensures the resultant reflection is considered. After the above inputs, hit Calculate Fields, wait for the computation to finish, hit Calculate Mechanical Adjustments, then hit Calculate Fields again. Inspect the resultant TUS focus with respect to the intended target using the two horizontal slider bars. If the focus is not acceptably close to the target, repeat the mechanical adjustment, calculate the field action until an acceptable focus-target distance (Figure 5A).
   NOTE: Refer to the TUS targeting software's documentation for transducer types other than Single.
9. Save the computationally optimized TUS target – example implementation: Close the TUS targeting software to return to the neuronavigation software. Notice that a new target Trajectory X-mod is added, where Trajectory X is the name given to the trajectory created in Step 3.5. This is the target for transducer placement during TUS. Close the Targets window and save the current project file.
10. NOTE: Steps 3.1 - 3.9 should be completed before the participant comes for TUS.

4. Non-simulation-based TUS targeting (participant scan optional, applicable to domains with readable TMS outcomes)

1. Initialize neuronavigated TMS - example implementation: Hit New Empty Project and Click here to choose an anatomical dataset to select participant MRI in the neuronavigation software. Hit New MNI Head Project if participant's structural MRI is not available.
   NOTE: Use a neuronavigation system/software of choice. This protocol uses a neuronavigation software at its 2.5.9 version (see Table of Materials) to illustrate representative procedures.
2. Skip this step if not using a participant MRI. Reconstruct the brain and scalp from participant MRI - example implementation: Go to the Reconstructions tab if using a participant MRI, then New Reconstruction… dropdown menu, select Skin, and push the Compute Skin button in the window opened to reconstruct the scalp. Close the skin window, New Reconstruction… dropdown menu, select Full Brain Curvilinear, set Slice spacing to 0.5 mm in the window opened, and hit Compute Curvilinear. After the brain is reconstructed, adjust the Peel depth slider bar until the reconstructed brain is exposed right at the outer surface of the cortex, and close the curvilinear brain window.
3. Skip this step if not using an MRI. Setup for physical head-MRI registration - example implementation: Go to the Landmarks tab if using an MRI, then Configure Landmarks…, and place landmarks on the nasion and left/right supratragic notch by single left click on the reconstructed scalp and hit New at the upper left. Confirm landmark placement by checking the crosshairs in the three MRI views. Adjust the crosshair(s) to fine tune the landmark placement, if needed. Close the registration landmarks window once satisfied and save the project file.
   NOTE: Steps 4.1 - 4.3 should be completed before the participant comes for TUS. Choose landmarks other than the nasion and the left/right supratragic indents as needed. Good landmarks are where the skin is not compressed/stretched during MRI acquisition.
4. Skip this step if TUS is to be used for the visual domain. Prepare for electromyogram (EMG) recording - example implementation: Identify the muscle represented at the intended TUS target, clean the skin overlying the muscle of interest with alcohol wipes, shave if needed and permitted, and apply EMG electrodes to the belly and tendon of the muscle. Apply another electrode to an adjacent bony area to act as the reference.
   NOTE: During skin preparation, exercise caution for participants on blood thinners.
5. Initialize neuronavigated TMS - example implementation: Go to the Sessions tab in the neuronavigation software, then New… dropdown menu, Online Session to launch a new session window. In this session window, go to the IOBox tab, under TTL Trigger options, check the correct channel/switch that receives trigger signal from the TMS stimulator, and ensure the Dead time: is set to 0 ms.
6. Skip this step if TUS is to be used for the visual domain. Setup integrated neuronavigation-EMG - example implementation: For motor domain, if using the neuronavigation system's EMG, check the correct Acquire EMG channel, and input a Trial: time sufficiently long to capture EMGs from the muscle of interest.
7. Prepare participant for neuronavigation - example implementation: Attach the subject tracker to the participant and ensure stability. Go to the Polaris tab in the neuronavigation software and position the camera/participant so the green dot (the participant) is roughly in the center of the three cyan boxes (viewing range of the camera).
8. Perform physical head-MRI registration - example implementation: Go to the Registration tab, single left click on a landmark, stably place the neuronavigation system's pointer on the respective landmark on the participant while ensuring both the pointer and the subject tracker are visible by the camera, and hit Sample & Go To Next Landmark. Repeat until all landmarks are sampled.
9. Skip this step if using a participant MRI. Scale template MRI to physical participant head - example implementation: If not using a participant's MRI, go to the Scaling tab, stably place the pointer on the scalp vertex of the participant, and hit Add Landmark. Repeat for the outermost point on the left, right, front, and back of the head.
10. Validate participant-MRI registration and update if needed - example implementation: Go to the Validation tab, gently slide the pointer on participant's scalp around the intended TUS/TMS entry point, and check for the Crosshairs <-> Skin distance, with closer to zero being better (< 1 mm can be used as a quantitative operational threshold). If the displayed distance is deemed unacceptable, go to Registration, Clear All, and repeat Steps 4.8 - 4.9 until an acceptable distance.
11. Setup the neuronavigation window - example implementation: Go to the Perform tab, bring up the Bullseye (Coil Centric), Curvilinear Brain & Samples, and Skin & Samples views, and set the Driver: to the correct TMS coil. If using the neuronavigation system's EMG, bring up the EMG and EMG Pod Recording too.
    NOTE: Use Bullseye (Target Centric) in place of Bullseye (Coil Centric) if desired.
12. Search for a TMS coil placement with responses - example implementation: Use both the curvilinear brain and the approximate position on the scalp to obtain initial starts, place the TMS coil at various locations and orientations on the scalp, and administer single TMS pulses over 10 s apart²³. If no response across the entire anatomically reasonable area, increase the stimulator output.
    NOTE: See Table of Materials for the TMS stimulator and coil models and EMG acquisition hardware and software. As an example implementation, the EMG acquisition had a 5000 Hz EMG sampling rate, 20 - 20000 Hz bandpass filter, 60 Hz notch filter, 1000x gain, manual artifact rejection, and used the MEP detection algorithm described in Soda et al., 2020²⁴.
13. Use a domain-specific response to define a target TMS coil placement - example implementation: Repeat Step 4.12 until a response is observed, i.e., EMG with pronounced peak-to-peak amplitude for the motor domain (Figure 6), and participants report of confident sight of phosphenes for the visual domain. Hit => Target and single left click this sample in the Targets to sample: list to make this coil location and placement a target.
14. Obtain additional repetitions at the initial TMS target - example implementation: Administer 2 - 3 additional pulses while aligning to this target by placing the coil in a way that the three metrics displayed in the Bullseye (Coil Centric) or the Bullseye (Target Centric) view are as close to zero as possible. Record the resultant responses.
15. Check for responses at placements adjacent to the initial TMS target - example implementation: Move the coil ~7 mm away from this target in the first quadrant in the Bullseye (Coil Centric) / Bullseye (Target Centric) view, fix the coil to this new location and orientation, administer ~3 pulses, and record the resultant responses. Repeat for the other three quadrants. If a non-center coil setup yielded a stronger average response than the center did, make the respective sample a target by single left-clicking it in the Recorded samples: list, => Target, and single left-click this sample in the Targets to sample: list.
16. Repeat Steps 4.14 and 4.15 until a center has the strongest average response among it and its four quadrants. This center's respective sample will be the target sample for TUS. Record this sample number and save the project file.

5. TUS setups

1. Connect TUS hardware and start oscilloscope recording - example implementation: Connect the function generators, amplifier, and the oscilloscope as illustrated in (Figure 7A). Turn on the oscilloscope, have it on the Run mode, and adjust the vertical and horizontal axes to the scale of interest.
   NOTE: Having two function generators is not always necessary. Select models of function generators can produce waveforms with more than one level of frequency control. Also, the hardware used in this protocol already matches in electrical impedances. Resistor(s) needs to be connected in series between the amplifier and the oscilloscope when electrical impedance is not a given.
2. Start and check the drive signal, then (for non-turnkey systems) start power output - example implementation: Flip the MAIN POWER switch of the amplifier from O to I. Dial the power output of the amplifier to zero, then configure the function generators to the waveform of choice, ensuring that ramping is added to the start and end of each PD. After checking the settings on both function generators, press Output on the top generator, then Output on the bottom generator, and then dial the power output of the amplifier to 100% by turning the ADJUST wheel and checking the displayed Gain: value, then press the POWER button of the amplifier to start its output (Figure 7B).
3. Check drive signal waveform and allow the system output to stabilize - example implementation: Confirm the waveform observed on the oscilloscope is as desired and let the drive system run for 15 min (empirically obtained criterion for hardware specified in Table of Materials) to ensure stable output (Figure 7C).
   NOTE: Stabilizing amplifier output is not related to voltage fluctuation. It is allowing the voltage to settle within the desired range, e.g., for the V_PP to decrease by 10-20% from the initial value and to maintain at that level, with the amplifier in this protocol (see Table of Materials). Different drive systems would have different V_PP decrease/increase and different durations to ensure a stable output. Determine empirically the system-specific behavior and duration required for stable output before having the first study participant.
4. Validate the ramping-based mitigation for auditory effects. Check the drive signal waveform at the start and end of each PD for ramping. Dial (or incrementally decrease) amplifier output to zero, fully immerse the transducer in water with the exit plane (i.e., stimulation side) ~1 inch below the water surface and facing up, then dial (or incrementally increase) amplifier output to 100%. Unsuccessful ramping implementation will cause water beads to be ejected from the water surface, resembling a fountain, whereas successful ramping will produce ripples on the water surface without beads leaving it.
5. Calibrate TUS transducer for neuronavigation - example implementation: Mount a neuronavigation tool tracker to the transducer holder. In the neuronavigation software, under Window, Tool Calibrations, hit New Calibration to create a new calibration for the TUS transducer. Input a name after Calibration name: and select the correct tool tracker at Tool tracker:. Hold the transducer mounted with the tool tracker against the calibration block, with the block's reference indicator pin touching and perpendicular to the center of the transducer's exit plane, hit Begin Calibration Countdown, hold still for 5 s, and confirm that the neuronavigation software gives Success. If not, repeat this last action until Success is seen. After completion, close Edit Calibration and Tool Calibrations windows.
   NOTE: For subsequent uses, replace the New Calibration action with single left click on the respective calibration under the Existing Tool Calibration list and hit Re-calibrate.
6. Terminate TUS setups - example implementation: Press the amplifier's POWER button (not its MAIN POWER switch) to turn off its output when the participant is ready for TUS, disconnect the oscilloscope from the amplifier, and connect the transducer to the amplifier output.

6. Administering TUS

1. Skip this step if using the non-simulation-based target. Load simulation-based TUS target to neuronavigation and perform participant-MRI registration (Steps 4.7-4.10) - example implementation: Go to the Sessions tab in the neuronavigation software if using the simulation-based target and the participant is ready for TUS, New… dropdown menu, and Online Session to launch a new session window. In there, under the Targets tab, single left click on the Trajectory X-mod target, and press Add => to make it a target for this session. Next, perform exactly Steps 4.7-4.10, then go to the Perform tab.
2. Setup neuronavigation for TUS administration - example implementation: Bring up at least the Bullseye (Coil Centric) or Bullseye (Target Centric) view in the Perform window, set the Driver: to the correct TUS transducer, and select the correct target in the Targets to sample: list on the upper left.
3. Prepare participant scalp for TUS coupling - example implementation: Comb the hair away to the extent possible, exposing the scalp at the approximate location for TUS transducer, and create an anterior-posterior parting line by doing so. Apply degassed ultrasound gel along this parting line, apply finger pressure center-out along the natural hair direction to burst air bubbles and saturate hair. Depending on transducer size, create 1-3 additional parting line(s) at ~1.5 cm distance on each side of the initial parting line and repeat the above actions. Next, return to the initial parting line and repeat the above actions 4 times.
   NOTE: This step adheres to the ITRUSST practical recommendation⁷.
4. Couple the TUS transducer to participant - example implementation: Strap the TUS transducer on the participant head and align the transducer to the target by adjusting its placement such that the three metrics displayed in Bullseye (Coil Centric) or Bullseye (Target Centric) are as close to zero as possible, while distance deviations < 1.5 mm, tilting deviations (pitch + roll) < 2 degrees, and rotational deviations (yaw) < 1 degree can be used as quantitative operational criteria. Once satisfied with transducer placement, slightly tighten the strap. Make sure the participant is comfortable and not choked.
5. (Optional but recommended) Insert a thermocouple wire in-between the transducer and scalp if online safety assessment is desired. Insert inside the ultrasound gel and ideally close to the scalp and not the transducer surface. Insert towards one edge of the transducer and not through the center such that the ultrasound wave is only minimally affected while having topical scalp temperature monitoring. Tape thermocouple wire on the head to keep it in place in case of head movement. Use paper tape for a balance between ease of removal and stickiness.
6. Start TUS and record initial transducer placement - example implementation: Press the amplifier's POWER button to turn on its output and thus the TUS. Simultaneously, press the Sample Now button at the lower left of the neuronavigation software window to record the initial transducer placement.
7. Record transducer placement throughout TUS and adjust as needed - example implementation: Hit Sample Now during TUS as desired to record transducer placement. Closely monitor the three metrics in Bullseye (Coil Centric) or Bullseye (Target Centric). Adjust the transducer placement if it deviates from the target, evident by the three metrics moving further away from zero (Figure 8).
8. Monitor the scalp temperature if a thermocouple is used and intervene if the thermocouple reading approaches 43.3 °C (110 °F)²⁵ or if the participant reports excessive heat sensation.
   NOTE: This 43.3 °C (110 °F) scalp threshold is to avoid ≥ 2^nd-degree scalp burns. The intracranial thermal dose limits are addressed by the ITRUSST safety guidelines¹⁸, stating that thermal effects are considered safe by ITRUSST if any of the three is satisfied: intracranial temperature rise ≤ 2 °C at any time, thermal dose ≤ 0.25 CEM43, or maximum exposure time depending on the level of thermal index (TI, 40 min for 2.0 < TI ≤ 2.5, and 10 min for 2.5 < TI ≤ 3.0). Check the ITRUSST safety guidelines for intracranial thermal calculations. TI usually means cranial TI (TIC) for most TUS setups¹⁸, but TI for soft tissue (TIS) and bone at focus (TIB) are also defined. Determine whether TIC, TIS, or TIB best corresponds to the scenario.

7. Finishing TUS

1. End TUS, check for participant wellbeing, and clean the participant's scalp - example implementation: Press the amplifier's POWER button to turn off its output and thus the TUS. Take off the transducer and all on-head equipment/setups, inspect the scalp for redness, and survey the participant for any discomfort or adverse event. After ensuring no discomfort or event, or after answering/resolving questions raised, gently wipe to clean the ultrasound gel to the extent possible.
2. End and export data of neuronavigation - example implementation: Hit Finish on the neuronavigation window and save the project by File and Save Project. Select the current session from the Sessions: list, hit Review, select Brainsight text file (.txt) from the Export… dropdown menu at the lower left of the window that opens, ensure Orientation (3 direction vectors) is selected and Snap to: dropdown menu is Nothing, configure other settings as desired, and hit Save to export the recorded TUS location and orientation during stimulation.
   NOTE: Ensuring no spaces in each sample name is not required, but a good practice.

8. Offline safety assessment (optional but recommended)

1. Acquire three-dimensional T1 magnetization-prepared rapid acquisition gradient-echo (MPRAGE) and a diffusion tensor imaging (DTI) sequence at the minimal. T2-weighted fluid-attenuated inversion recovery (FLAIR) images at the same resolution as the T1 MPRAGE will be optional but valuable, especially for stroke. Acquire pre- and post-TUS.
   NOTE: Choose neuroimaging parameters and scanner as desired, this protocol has T1 MPRAGE: 1 mm isotropic spatial resolution, 2400 ms TR, 3 ms TE, and 900 ms TI; FLAIR: 1 mm isotropic spatial resolution, 5000 ms TR, 90 ms TE, and 1806 ms TI; and DTI: dual spin-echo, echo-planar imaging acquisitions, 2 mm isotropic spatial resolution, b factor = 1000 s/mm², and 2.5 mm slice thickness. See the Table of Materials for the scanner used.
2. Convert raw images to NIFTI (i.e., .nii format), if not defaulted so, using a method of choice. See the Table of Materials for example methods.
3. Separate the 4D DTI image and obtain a brain mask - example implementation: Install the DTI processing software²⁶ if not already installed (see the Table of Materials), run fslsplit on the 4D diffusion NIFTI, and run bet on the resultant vol0000 file for a brain mask²⁷.
4. Correct for eddy current-induced distortions in the DTI image - example implementation: Create an index.txt file of a single row of N 1s with a single space in between (e.g., for N = 5, index.txt would contain only: 1 1 1 1 1) where N is the number of values in the bval file, and run eddy²⁸ with --imain, --mask, --bvecs, --bvals, --index, --acqp, --data_is_shelled, --repol, and --out options, where the index text file is the input to --index, and a single row of space-delineated text file containing 0, 1, 0, 0.08 (i.e., 0 1 0 0.08) is the input to --acqp.
5. Skip this step for neurotypical adults. Delineate the lesion for a lesion mask - example implementation: Use a lesion segmentation tool of choice or draw manually on FLAIR a mask of the lesion if studying a cohort with lesion, use T1 if FALIR is not available, and save the mask as a NIFTI.
6. Segment the anatomical MRI images - example implementation: Install the MRI processing software (see Table of Materials) if not already installed, run spm fmri in the programming platform (see Table of Materials) command window, hit the Segment button, SPM dropdown menu, Tools, Clinical, and MR segment-normalize to bring up the segmentation window. Input T1 to Anatomicals, lesion mask to Lesion maps (leave blank if neurotypical), FLAIR to Pathological scans, and hit the green triangle to start the segmentation²⁹.
7. Fit a diffusion tensor model to the eddy-corrected DTI image for a scalar DTI map - example implementation: Run dtifit³⁰ with the --data, --out, --mask, --bvecs, --bvals, and --verbose options on, where entry to --out in Step 8.4 (the eddy step) is the input to --data. Run fslsplit on the 4D diffusion image post dtifit to create a new vol0000.
8. Obtain a representative transducer location and orientation during TUS administration - example implementation: Use custom script or published pipeline³¹ on the text file from the neuronavigation software (Step 7.2), import this representation in the neuronavigation software by Targets tab, New… dropdown menu upper left, and Import from File… the representation, single left click on it under the Name list, and hit Go to to confirm that it is on a sensible location and orientation.
   NOTE: Ensure this neuronavigation software file is built on the pre-TUS MRI image. Ensure the neuronavigation software-derived transducer location and orientation have coordinates in the same space as the MRI image on which the acoustic simulation is run.
9. Obtain a region of interest (ROI) for TUS-stimulated tissue - example implementation: Run an acoustic simulation by performing Steps 3.1 - 3.2 and 3.6 - 3.8. Run Calculate Fields once during Step 3.8.
10. Align pre-TUS anatomical scans and tissue segment masks from Step 8.6 to the pre-TUS, b0 scalar DTI map (from Step 8.7) - example implementation: Use the MRI processing software's coregister-est&reselice where the vol0000 is the reference, pre-TUS T1 is the source, and the c1 to c5 files from Step 8.6 for pre-TUS are input to other images. Align post-TUS, dtifit-ed vol0000 to pre-TUS, dtifit-ed vol0000 using the same tool (i.e., reference = pre-TUS, dtifited vol0000 and source = post-TUS, dtifit-ed vol0000), then align post-TUS T1 and segment masks from Step 8.6 to either of the vol0000 similarly.
11. Intersect the TUS-stimulated ROI (Step 8.9) with the gray- and white-matter masks for the final ROI for ADC reading - example implementation: Binarize gray matter and white matter masks by running fslmaths rc1_* -thr 0.5 -bin ID_GM.nii and fslmaths rc2_* -thr 0.5 -bin ID_WM.nii, combine the resultant masks by running fslmaths ID_GM.nii -add ID_WM.nii ID_WGM.nii, and obtain the intersection between this combined mask and the simulated acoustic field by running fslmaths *_FullElasticSolution.nii.gz -mas ID_WGM.nii ID_roi_WGM.nii where *_FullElasticSolution.nii.gz is the simulation output from Step 8.9.
12. Overlay the final ROI mask (Step 8.11) and obtain pre- and post-TUS ADCs - example implementation: Overlay this ID_roi_WGM.nii to the pre- and post-TUS dtifit-ed vol0000 in the MRI viewer (see Table of Materials), then Draw, Descriptive, record the value at MEDIAN, and obtain the post-vs-pre ADC change by computing the (post-pre)/pre percentage using the pre- and post-TUS medians.

9. Assessing CSE variations induced by TUS

The steps in this section are modular. Perform them to assess CSE at the desired frequency and timepoints relative to TUS. Choose the frequency and time points based on the specific research question. See Table of Materials for the TMS stimulator and coil models and EMG acquisition hardware.

1. Assess CSE on the muscle whose cortical representation receives TUS and perform Step 4.4 to this muscle for EMG electrodes to be prepared.
2. Perform Steps 4.5-4.11 to prepare neuronavigation.
3. Determine the TMS coil setup (placement and orientation) that best stimulates the muscle of interest, i.e., the hotspot.
   1. Use the TMS hotspot obtained during non-simulation-based TUS targeting if the TUS targeting is non-simulation-based - example implementation: on the neuronavigation software, in the list below Targets to sample:, single left click the sample that yielded the strongest EMG response to make it a neuronavigation target.
      NOTE: See Table of Materials for the TMS stimulator and coil models and EMG acquisition hardware and software. As an example implementation, the EMG acquisition had a 5000 Hz EMG sampling rate, 20 - 20000 Hz bandpass filter, 60 Hz notch filter, 1000x gain, manual artifact rejection, and used the MEP detection algorithm described in Soda et al., 2020²⁴.
   2. If the TUS targeting is simulation-based, perform Steps 4.12 - 4.16 to identify the hotspot.
4. Determine the 1 mV motor threshold (MT_1mV) - example implementation: Place the TMS coil aligning to the hotspot obtained above by minimizing the three metrics displayed in the bullseye window (distance deviations < 1.5 mm, tilting deviations (pitch + roll) < 2 degrees, and rotational deviations (yaw) < 1 degree), determine the TMS stimulator output needed to elicit 1 mV MEPs, i.e., MT_1mV, using an algorithm of choice (e.g., PEST³²). If 1 mV MEP is not attainable at 100% maximum stimulator output (MSO), determine the percentage MSO needed for 0.05 mV MEPs.
   NOTE: The percentage MSO needed for 0.05 mV MEPs is not MT_1mV (see Step 9.5). Also, when performing this section more than once, this MT_1mV step only needs to be determined in the first time, unless MT_1mV itself is chosen to be used as a global measure of CSE.
5. Deliver 20 single TMS pulses at the MT_1mV with an inter-pulse interval > 10 s²³.
   NOTE: If MT_1mV is not attainable but MT_0.05mV is, use 100% MSO. If both MT_1mV and MT_0.05mV are not attainable, this participant would be ineligible for this CSE assessment.

10. Assessing MSL variations induced by TUS

NOTE: The steps in this section are modular. Perform them to assess MSL at the desired frequency and timepoints relative to TUS. Choose the frequency and time points based on the specific research question.

1. Use software/language of choice (see Table of Materials for examples) to build a discrete sequence production task where a five-button physical keyboard is placed in front of a computer screen that shows five horizontally equidistant, 2 cm x 2 cm empty boxes, aligned with the five keyboard buttons. Have a bead to appear in one box at a time, disappear when the box's corresponding button is pressed, and appear in another box. Have the bead to appear in the five boxes without repetition, forming a sequence of appearance, for example, box number 1-4-3-2-5. Have the task repeat a sequence twelve times. Have a welcome page before the task starts and have the transition from this welcome page to the task initiated by a press of any key. Have a response time recorder that tracks the time elapsed between each bead appearance and the corresponding button press. An example pseudocode would be
   boxes = [1, 2, 3, 4, 5]
sequence = [1, 4, 3, 2, 5]
repeats = 12
rt_log = []
print("Welcome! Press any key to begin.")
wait_for_any_key()
show_boxes() # draw empty boxes
for r in range(repeats):
  for target in sequence:
    show_bead(target) # bead in current box
    start = current_time()
    while True:
      key = wait_for_key()
      if key == target: # correct button
        rt = current_time() - start
        rt_log.append(rt)
        clear_bead()
        break
      else:
        continue # ignore wrong presses
print("Task complete. Thank you!")
save(rt_log, "results.csv")
   NOTE: The number of buttons/boxes and sequence repetition can vary depending on the experiment design. Considerations are the level of function in the cohort of interest, and the balance of sequences across participants, i.e., number of participants should be divisible by the number of possible sequences. The interval between bead disappearance and the next appearance, and between end of one sequence to the start of the next sequence, is also experiment design dependent.
2. Have the participant sitting with comfortable posture. Place the keyboard at a location of comfortable reach with the paretic hand of the participant.
3. Explain to the participant that, "There will be five boxes on the screen. A bead/circle will appear in one of the boxes. When you see the bead/circle, press the corresponding key using your stroke-affected hand. Do this as fast as you can as we will be timing, okay? Any questions?"
   NOTE: Avoid the mention of "sequence" in any form. MSL needs to be implicit learning for it to be a measure of the state of the motor cortex³³. The mention of "sequence" increases the chance of participants indexing each box and explicitly remembering the sequence, voiding MSL's ability to assess the state of the motor cortex.
4. Assign a sequence randomly to the participant after answering any questions the participant may have, let them know that "You can press any key to start whenever you are ready."
5. Monitor participant's posture during the task and intervene if any sign(s) of a fall. At the same time, if the participant uses the other hand/arm to help, remind them that only the paretic hand should be used for this task.
6. Check if the participant is well at task completion, remove the keyboard/screen, praise the participant for their performance, and express gratitude.

11. Data analysis

NOTE: The overall data analysis is dependent on the experimental design. Steps below provide instructions for one basic unit of CSE/MSL measure together with a simple, canonical example of percentage post-pre comparison.

1. CSE analysis
   1. Check each EMG trace for an identifiable MEP that has an onset latency of at least 19 ms (for APB), calculate the peak-to-peak amplitude between the onset and offset of each MEP, and calculate the median peak-to-peak amplitude for the 20 MEPs within the block. This median amplitude will be the measure of CSE for this block.
      NOTE: The onset of MEP latency depends on the muscle, muscle state (rest vs. active), and neurological disorder, e.g., upper extremity: 10-40 ms³⁴; lower extremity: 30-45 ms^35,36. Consult with the literature if the muscle of interest is not APB. Onset/offset of a MEP can be determined using the thresholding method described in Soda et al., 2020²⁴.
   2. Obtain the percentage post-pre CSE change by subtracting the post CSE measure from the pre CSE measure and dividing this difference by the pre CSE measure.
2. MSL analysis
   1. Calculate the median response time within each sequence repetition, i.e., median of the five button presses, then calculate the median of the twelve within-repetition medians. This "median of median" will be the measure of MSL for this block.
   2. Obtain the percentage post-pre MSL change by subtracting the post MSL measure from the pre MSL measure and dividing this difference by the pre MSL measure.

Simulation-based TUS targeting first defines an initial transducer placement manually based on proximity to the intracranial target and perpendicularity to the skull and uses this initial placement and transducer specifications to solve for the intracranial acoustic field. Next, it computes the discrepancy between the intended stimulation target and the acoustic field focus, updates the transducer placement in the reverse direction of this discrepancy, solves for the acoustic field using this new transducer placement, and iterates until the acoustic focus reaches the intended stimulation target to the extent possible. Figure 5A shows an example of satisfactory targeting to a cortical region, and Figure 5B exemplifies a scenario where additional iterations would help. Figure 5C underscores an unexpectedly important entry, the Distance Tx outplane to skin (mm). The TUS targeting software defaults this distance to values larger than typical transducer-scalp distances, causing the results window to be illegibly sized. Inputting a more truthful transducer-scalp distance solves this issue and gives desirable results window sizes (Figure 5A-B).

Non-simulation-based targeting uses TMS-elicited MEPs to determine the scalp location for transducer placement and has perpendicular-to-scalp as transducer orientation. Figure 6A shows a clean MEP that indicates good TMS coil placement and orientation. Figure 6C shows a possibly non-MEP muscle activity, exercise cautious and continue with the hotspot search (Steps 4.12-4.16). Step 9 also uses TMS-elicited MEPs, though for the different purpose of assessing the CSE, so Figures 6A, 6C are also representative results for Step 9. In addition, Step 9 includes MT1mV determination (Step 9.4). Figure 6B represents a clear MT1mV determination of 52% MSO as the positive and negative MEP statuses were even at 52%: rounding the second column to integers, 52% had five 1.0 s and five 0.0 s in the third column. Figure 6D shows a potentially inconclusive MT1mV determination and prompts a MT1mV retest or a new coil placement/orientation, even though PEST suggests 87%: consecutive yeses (1.0 in the third column) and noes (0.0 in the third column), while an alternating yeses and noes around the same MSO is the ideal case.

Figure 7C shows a steady, even oscilloscope reading of the transducer driving signal. Suboptimal reading may have baseline drifts, insufficient peak-to-peak amplitude, or arbitrary overshoots, in which case repeat Step 5.3 until steady, even readings. Suboptimal readings could also mean equipment (function generator, amplifier, etc.) wear and tear.

Figure 8A shows a good alignment between the physical transducer placement during TUS and the pre-defined target, while Figure 8B shows when the transducer drifted and requires adjustments. Note in the Skin & Samples view how indifferentiable an off-target transducer placement (the mini TMS coil) would be compared to the intended transducer placement (the red lines). While TUS can be administered without neuronavigation, this observation underscores the importance of using neuronavigation.

In Huang et al., 20252, from which this protocol is derived, the feasibility and safety of such TUS administration to stroke participants is demonstrated, as eighteen stroke participants completed all study activities with zero occurrences of predefined adverse events (≥ 2nd-degree scalp burn, clinical seizure, new lesion on DTI sequence or major reduction of ADC, and participant withdrawal request due to any reason). Indications of efficacy were also reported, providing a proxy for the accuracy of the protocol. A significantly greater proportion of participants in the high-dose group (4, 6, and 8 W/cm2 estimated in vivo spatial-peak pulse-average intensity, ISPPA) showed ≥ 20% improvement in MSL response time after TUS compared with the low-dose group (0, 1, and 2 W/cm2 ISPPA), 67% (6/9) versus 0% (0/9), p = 0.009.

Figure 3: TUS components and targeting decision flowchart. (A) Overview of the components for transcranial ultrasonic stimulation (TUS) in clinical cohorts. (B) Decision flowchart to determine between simulation-based (Step 3) and non-simulation-based (Step 4) TUS targeting. Please click here to view a larger version of this figure.

Figure 4: Setup of transducer characterization. Experiment setup of ultrasound transducer characterization using a hydrophone, adapted from Chen et al., 202314. RF: radio frequency. Please click here to view a larger version of this figure.

Figure 5: Transcranial ultrasonic stimulation (TUS) targeting by comparing the simulated acoustic field (the heatmaps) and intended stimulation target (the black cross). (A) Desirable alignment of the simulated TUS focus and the intended target. (B) Suboptimal alignment of the simulated TUS focus and the intended target from insufficient iterations of Calculate Mechanical Adjustment (Step 3.8). (C) Illegibly sized results window due to an incorrectly large transducer-scalp distance (red box). Please click here to view a larger version of this figure.

Figure 6: Motor-evoked potentials (MEPs) from transcranial magnetic stimulation to the motor cortex and motor threshold determination procedure based on parameter estimation (PEST). (A) Salient MEP at 24 ms post-stimulus (0 ms) with low background and pre-stimulus noise. (B) Balanced ending of PEST with even distribution of positive and negative MEP status at the ending MSO of 52%. (C) Questionable MEP at 20 ms post-stimulus riding on top of a noisy background and pre-stimulus activities. (D) Equivocal ending of PEST with uneven numbers of positive and negative MEP determination status at the ending MSO of 87%. Please click here to view a larger version of this figure.

Figure 7: Hardware setups for transcranial ultrasonic stimulation (TUS). (A) Schematic diagram of essential TUS hardware and its connections. (B) Control panel of the amplifier, on which the POWER button (and not the MAIN POWER switch at the back) should be used to turn on/off output throughout the TUS visit. (C) Steady, even oscilloscope readings indicating the drive system is ready for administrating TUS. Please click here to view a larger version of this figure.

Figure 8: Representative location and orientation tracking for the transducer via neuronavigation. (A) Low transducer placement and orientation error as measured by neuronavigation. (B) Large transducer placement error as seen by neuronavigation. Please click here to view a larger version of this figure.

Table 1: Example TUS parameters. 1.1 Pulse timing parameters in Huang et al., 20252 following the ITRUSST reporting guidelines10. 1.2 Additional parameters in Huang et al., 20252. ISPPA, spatial-peak pulse-average intensity; ISPTA, spatial-peak temporal-average intensity.

Supplementary Material 1: Sample pre-TUS checklists for eligibility for consent, safety screening, and TUS equipment quality control. Please click here to download this File.

Supplementary Material 2: Sample post-TUS checklist for adverse event monitoring. Please click here to download this File.

There has been significant research enthusiasm for the application of TUS in neurological conditions; however, a need exists for increased scientific rigor and replication of results. A successful TUS on the intended target tissue relies on several critical steps. First, the transmitting sensitivity of the transducer needs to be correctly characterized. Options include following Step 2 of this protocol, outsourcing to a collaborator, and utilizing vendor's service. The second critical step is targeting, i.e., determining the transducer location and orientation on the scalp. For shallow and broad target tissue, TMS-based motor cortex targeting (Step 4) has been shown to have an influence on corticospinal excitability as well as behaviors^2,3. For deep (lower cortical, subcortical) and smaller target tissue, or for relatively more precise targeting in general, simulation-based targeting (Step 3) is the current state-of-the-art in TUS⁶. Third, it is often neglected that the amplifier and transducer need to have a "warm-up" first to provide stable output (Steps 2.4 and 5.3). Fourth, careful attention must be given to preparing the scalp for TUS (Step 6.3), as a quick back-of-the-envelope calculation shows that even a small air bubble can reflect over 80% of the TUS energy. Fifth, close monitoring during TUS (Steps 6.7-6.8) is essential to ensure accurate stimulation to the target tissue and maintain participant safety, especially given the early stage of TUS research.

Proper electrical impedance matching between the drive system and the transducer is critical for efficient power transfer and the prevention of energy loss through electrical reflections. Standard coaxial cables connecting the drive system (separate or turnkey) to the transducer typically have a 50 Ω electrical impedance. Transducers are usually supplied with an impedance matching unit, but this should be confirmed with the vendor. If impedance matching is not provided, impedance matching resistor(s) should be connected between the drive system and the transducer to optimize power transfer. Even when impedance matching is ensured, a power meter is valuable for confirming the actual TUS output, adding safety assurance and scientific rigor. Similarly, if the function generator(s) or the turnkey drive system cannot log the output waveform, placing an oscilloscope between the amplifier and the transducer allows real-time validation and an offline record of the driving signal. A neuronavigation system enables spatial localization of the transducer in relation to the participant during TUS. Optical neuronavigation employs infrared fiducials (i.e., infrared-reflective spheres about half an inch in diameter attached to both the transducer and the participant), along with an infrared camera. The previously mentioned 3D-printed transducer mount can be designed with openings to secure these fiducials, as exemplified by the mount in Figure 1.

The protocol can be modified to include an alternative method for minimizing the experimental confound of auditory effects. In place of the primary mitigation method (adding ramping to the chosen waveform, Steps 5.2 and 5.4), participants can wear a bone-conduction earphone that plays a tone matching the spectral peak of the TUS waveform during stimulation (i.e., during Steps 6.6-7.1). Bone-conduction earphones (rather than conventional air-conduction earphones) are currently standard practice because the auditory effect of TUS is believed to arise from skull-conducted mechanical waves³⁷, rather than airborne pressure waves. At Step 8.8, one could encounter difficulties ensuring the neuronavigation software-derived transducer location and orientation have coordinates in the same space as the MRI image for acoustic simulation. To troubleshoot, one can view the transducer location on an MRI viewer of choice (see examples in Table of Materials), visually find this location in the neuronavigation software slice-by-slice (Sessions tab, New… dropdown menu, Offline Session, Perform tab), single left click on the MRI image and adjust the AP, Lat, and Twist sliders (see Step 3.5) to have the transducer normal to that scalp location, and create a target in the neuronavigation software by hitting Sample Now. Next, make this sample a target, close the perform window, go to the Targets tab, Configure Targets…, then continue to Step 8.9 for the acoustic simulation.

The current targeting approaches represent one key limitation in TUS practices, as numerically solving for the viscoelasticity in TUS scenarios is still an active area of research. For instance, the most widely used TUS simulation pipelines employ a Cartesian computational grid such as the finite-difference time-domain²¹, which is susceptible to the staircasing effect³⁸. This effect becomes more pronounced when using data-driven skull models derived from MRI or CT scans, easily leading to errors in both the location and amplitude of the simulated TUS focus. Another major bottleneck lies in the design of TUS waveforms, specifically determining parameters such as RD, PD, PRI, and PRD. Currently, there is no first principles method for parameterizing the driving signal's pulse shape, forcing investigators to rely on empirically deriving parameters from previous studies to design their TUS waveforms.

Nevertheless, these limitations are common across current experimental human TUS studies. This protocol offers step-by-step guidance for applying TUS to neurologically diseased populations, particularly stroke patients. By promoting standardization in TUS clinical research, it aims to accelerate the translation of this technology into clinical practices. Additionally, this protocol introduces a minimalistic on-head fixation method that enhances participant movement, clinical usability, and supports total stimulation times exceeding 10 min, significantly longer than the typical protocols (80 s) used in neurotypical adults. This extended stimulation window and improved tolerability open opportunities for testing repeated-session interventions, dose-response studies, and combined TUS-rehabilitation paradigms in stroke recovery. Furthermore, the ability to tolerate naturalistic movement (within hardware tethering) enables designs that assess TUS effects during ambulatory tasks, motor training, or interactive rehabilitation settings, broadening its potential applications beyond static, seated experiments.