Transcranial ultrasound stimulation (TUS) is an emerging non-invasive neuromodulation technique that requires careful planning of acoustic and thermal simulations. The methodology describes an image processing and ultrasound simulation pipeline for efficient, user-friendly, streamlined planning for human TUS experimentation.
Transcranial ultrasound stimulation (TUS) is an emerging non-invasive neuromodulation technique capable of manipulating both cortical and subcortical structures with high precision. Conducting experiments involving humans necessitates careful planning of acoustic and thermal simulations. This planning is essential to adjust for bone interference with the ultrasound beam’s shape and trajectory and to ensure TUS parameters meet safety requirements. T1- and T2-weighted, along with zero-time echo (ZTE) magnetic resonance imaging (MRI) scans with 1 mm isotropic resolution, are acquired (alternatively computed tomography x-ray (CT) scans) for skull reconstruction and simulations. Target and trajectory mapping are performed using a neuronavigational platform. SimNIBS is used for the initial segmentation of the skull, skin, and brain tissues. Simulation of TUS is carried over with the BabelBrain tool, which uses the ZTE scan to produce synthetic CT images of the skull to be converted into acoustic properties. We use a phased array ultrasound transducer with electrical steering capabilities. Z-steering is adjusted to ensure that the target depth is reached. Other transducer configurations are also supported in the planning tool. Thermal simulations are run to ensure temperature and mechanical index requirements are within the acoustic guidelines for TUS in human subjects as recommended by the FDA. During TUS delivery sessions, a mechanical arm assists in the movement of the transducer to the required location using a frameless stereotactic localization system.
Commonly used non-invasive neurostimulation techniques include transcranial direct current stimulation (tDCS) and transcranial magnetic stimulation (TMS). However, both have limited penetration depth and low precision1,2. By contrast, transcranial ultrasound (TUS) is an emerging non-invasive technique capable of enhancing or suppressing neuronal activity3,4,5 and targeting cortical or subcortical structures at millimeter precision6,7. Animal models using rodents4,8,9, rabbits10, sheep5,11, swine6, and nonhuman primates7,12,13,14 have shown the efficacy and safety of TUS. Studies have demonstrated that targeting various brain regions can elicit limb movements8 in rats, somatosensory evoked potentials (SSEPs) in swine6, and changes in visuomotor activity12, cognitive, and motivational decision-making in nonhuman primates13 among other changes in behavior. In humans, TUS has been observed to change motor evoked potentials (MEPs) and performance on a reaction time task when targeting the primary motor cortex15,16 and altered performance on a tactile discrimination task and SSEPs when targeting the somatosensory cortex17 and sensory thalamus18. Histological analyses have revealed no gross or microscopic structural changes associated with TUS in swine6, sheep5,11, rabbits10, and nonhuman primates14, and no side effects have been seen that significantly differ from other non-invasive neurostimulation techniques19.
TUS uses pulsed low-intensity focused ultrasound at a frequency between 200 kHz and 700 kHz to produce a transient neuromodulatory effect. The typical spatial-peak pulse-average intensity (Isppa) in situ is 10 W/cm2 or less, with reported duty cycles (percentage of time when ultrasound is on) ranging from 0.5% to 70% in humans20,21,22,23,24. Although the mechanisms of TUS neuromodulation have been proposed to mainly involve mechanical agitation of lipid membranes leading to the opening of ion channels25,26,27, possible thermal and cavitation effects cannot be ignored. They are assessed through mechanical (MI) and thermal (TI) indices. The MI describes the predicted cavitation-related bioeffects that will occur with TUS, whereas the TI describes the potential temperature increase within tissues following ultrasound application28,29. Furthermore, changing the frequency and input intensity also causes the MI and TI to change. Higher frequencies have better spatial resolution and decrease the probability of mechanical bio-effects; however, they have stronger absorption in the tissue, which increases the potential for temperature rise28. Alternatively, lower frequencies at the same intensity increase the MI. Similarly, increasing the intensity tends to increase the magnitude of mechanical and thermal bio-effects30. It is, therefore, imperative that careful planning and simulation be performed before experimentation sessions for all TUS parameters that will be implemented.
Planning a TUS experiment requires the identification of the target and trajectory of interest and the performance of thermal and acoustic simulations. Simulations assist in optimizing mechanical effects and mitigating the thermal effects of TUS on tissues. They require understanding the prediction of skull heating, pressure amplitude of the ultrasound at the focal point, focal correction, and other heating within the skull and skin. Adequate simulation ensures the focal point will reach the target of interest and safety parameters for ultrasound use set out by the safety guidelines on biophysical safety as recommended by the International Transcranial Ultrasonic Stimulation Safety and Standards Consortium (ITRUSST)31, which are based on FDA and Health Canada recommendations, are followed. Recent studies have also highlighted an auditory confounding effect accompanied by TUS32,33,34 in animals and humans, whereby TUS stimulation can activate auditory pathways in the brain to elicit responses32,33,34. Transection of the auditory nerves32, removal of cochlear fluid32, or chemical deafness33 in rodents have been employed to diminish these effects in animals. In humans, administering an auditory tone through headphones has been used to effectively mask auditory noise from TUS, controlling for the TUS-induced auditory activity confound34. This highlights the need to control for auditory noise in sham stimulation conditions, which must be incorporated into protocol planning, design, and implementation.
Here, we present a guide on how to appropriately complete the preparation (step 1, step 2), planning (step 3), simulations (step 4), and TUS delivery (step 5) required to perform TUS neuromodulation experiment in humans.
All methods involving the use of human subjects were performed in compliance with the Tri-Council Ethical Conduct for Research Involving Humans, and the protocol was approved by the Conjoint Health Research Ethics Board (CHREB) at the University of Calgary. All subjects provided informed written consent before participation. Human participants were required to be healthy, right-handed adults between ages 18 and 40 willing and able to complete a magnetic resonance imaging (MRI) scan. Exclusion criteria included family history of seizure, mood, or cardiovascular disorders, ear trauma, alcohol or drug dependency, use of prescription medications, metal implants including a pacemaker, pregnancy, cardiovascular disorders, history of neurological or psychiatric disorder, inability to communicate with investigator and study staff, and legal incapacity or limited legal capacity. The protocol described below follows the recommendations by ITRUSST on the standardized reporting of TUS studies35. The details of the equipment, software, and necessary weblinks used in this study are listed in the Table of Materials.
1. High-resolution magnetic resonance imaging
2. Pre-processing participant images
3. Trajectory planning
Figure 1: Creating a full brain curvilinear in Brainsight. (A) Box adjusted to the edge of the sagittal MR image. (B) Box adjusted to the edge of the coronal MR image. (C) Box adjusted to the edge of the transverse MR image. (D) Full brain curvilinear reconstruction with a peel depth of 4 mm. Please click here to view a larger version of this figure.
Figure 2: Landmarks placed on the skin reconstruction and MR image. (A) Nose and nasion landmarks placement. (B) Left ear landmark placement. (C) Right ear landmark placement. Please click here to view a larger version of this figure.
4. Simulations with BabelBrain
NOTE: Details for simulation with BabelBrain can be consulted in BabelBrain's manual: https://proteusmrighifu.github.io/BabelBrain/index.html.
Figure 3: Acoustic simulation using BabelBrain. Please click here to view a larger version of this figure.
Figure 4: Thermal simulation using BabelBrain. Please click here to view a larger version of this figure.
5. TUS delivery session
Figure 5: Fiducials to be used for neuronavigation. Glasses (left) and headband (right) with fiducials attached for subject tracking. The subject's head is secured with a chin rest and stabilizer behind their head. Please click here to view a larger version of this figure.
Figure 6: Brainsight neuronavigation screen for target localization during TUS delivery session for a primary motor cortex (M1) target. Rotational and translational indicators guide the experimenter on where to position the transducer over the scalp and when the trajectory angle has been achieved. Accuracy to target indicates how close the translational and rotational orientation is to the target and should be used to fine-tune movement. The depth indicator shows the depth of the focal spot and should be used to lower the transducer to the appropriate height. Please click here to view a larger version of this figure.
Figure 7 illustrates comparative session samples from one of our studies42, featuring two distinct participants employing specific ultrasound parameters (fundamental frequency of 250 kHz, sonication duration of 120 s, a pulse repetition frequency (PRF) of 100 Hz, a duty cycle of 10%, and an ISPPA of 5 W/cm²). In this research, T1-, T2-w, and ZTE MRI scans with 1 mm isotropic resolution were obtained from neurologically healthy subjects. TMS targeted the primary motor cortex (M1) to assess the impact of varied TUS PRFs (10 Hz, 100 Hz, and 1000 Hz) on corticospinal excitability at distinct time points pre – and post-TUS (30 single-pulse TMS measurement at each time point as indicated by the individual points in the figure). Figure 7 encompasses results from a successful experiment, displaying a prolonged inhibitory effect following TUS (Figure 7A), alongside data from a participant where a significant effect was not observed (Figure 7B). Previous TUS studies have reported that some participants do not show any significant effects associated with ultrasound stimulation45,46. Similar results have been observed in studies using TMS47,48 and tDCS49,50, showing that NIBS techniques do not always result in expected effects. This highlights the necessity of carefully planning and implementing TUS experiments in humans using protocols such as the one described to maximize the chance of achieving experimental success.
The study by Zadeh et al.42 accentuates the differential effects of TUS, showing that 10 Hz and 100 Hz PRFs can induce prolonged inhibitory effects, unlike the negligible impact observed with a 1000 Hz PRF. Importantly, the research demonstrates that a 100 Hz PRF is most effective, achieving an extended suppression of MEPs exceeding 60 min, in contrast to the 30 min suppression achieved with a 10 Hz PRF. Alternative methods of assessing the effect of TUS may be employed, such as measuring SSEPs or task performance, each requiring a unique interpretation of results.
Figure 7: Transcranial magnetic stimulation (TMS) targeting the primary motor cortex (M1) to assess the impact of varied TUS pulse repetition frequencies (PRF) on corticospinal excitability. TUS was delivered at a fundamental frequency of 250 kHz, sonication duration of 120 s, pulse repetition frequency (PRF) of 100 Hz, a duty cycle of 10%, and an ISPPA of 5 W/cm2. Thirty trials at each time point were recorded, as indicated by the individual dots. The results presented highlight the differences in the median normalized MEP amplitude over 5 min, 30 min, and 60 min after TUS exposure. Error bars show the interquartile range for trials at each time point. (A) Subject showing successful experimental results exemplifying prolonged inhibitory effects. (B) Subject showing no significant effect. Please click here to view a larger version of this figure.
In this method, subject-specific simulations are performed to predict and assess possible thermal and mechanical effects resulting from TUS application to the brain. Data sets between participants must remain separate and carefully documented, as using an incorrect scan or data file will lead to inaccurate simulations. When numerous participant scans are collected, and planning is performed together, it is important to ensure proper labeling of images and folders and proceed with caution when sorting and saving files.
To accurately perform simulations, the skull must be correctly modeled to gain a subject-specific understanding of acoustic and geometric properties. BabelBrain, therefore, requires the completion of a CT or ZTE scan; if neither is available, BabelBrain can use a simplified mask of the skull bone obtained with the charm tool. CT scans provide the most accurate information about skull characteristics to predict how the ultrasound will propagate in the brain once it contacts the skull. This is achieved by using voxelized information of the skull bone, although estimation of bone volume is still difficult to predict due to its heterogeneous nature51. However, more recently, ultra-short TE MRI sequences, such as the ZTE scan, have been used to determine the acoustic properties of the skull and perform TUS simulations. Bone pixel resolution and contrast are not as high in a ZTE image as they are in CT; however, compared to simplified masks derived from just T1-w and T2-w MRI, ZTE images still provide an improved estimation of heating with TUS simulations52 having predictions similar to those made with CT scans53. Using ZTE scans in simulations eliminates any unnecessary exposure to radiation and additional time required to obtain X-ray CT images, as ZTE sequences can be added to the original MRI scanning session. However, ZTE protocols are not a standard product on most MRI scanners, and additional research is required to improve ZTE-based post-processing for use in TUS modeling. If neither CT nor ZTE images are available, the charm tool utilized with BabelBrain creates a simplified mask from T1- and T2-w images to predict skull characteristics. A comparison of simulations performed with a CT versus a simplified mask has seen minimal differences in steering (1.00 mm ± 1.3 mm) and mechanical repositioning (x-direction 0.2 mm ± 0.4 mm; y-direction 0.1 mm ± 0.5 mm) required to reach the target, however slightly different intensity distribution (focus length 6% ± 8%; focus width 2.7% ± 2%; focus volume 10.4% ± 16%; peak pressure -0.5% ± 15% for simplified mask relative to CT)36. This reveals that the CT provides more accurate simulations; however, if unable to utilize a CT or ZTE scan, the simplified mask is adequate to complete simulations for TUS experimentation in humans.
Mapping methods of acoustic properties remain a necessary area of research, as prediction of the speed of sound and attenuation from CT and MR images still require a better understanding and streamlined workflow. Therefore, a caveat of potential deviation in thermal or mechanical effects from what is predicted in the simulation remains. Modeling of acoustic effects in BabelBrain is achieved using an open-source BabelViscoFDTD54,55 library, which allows modeling of both linear compressional and shear wave transmission. This model has been validated against multiple other models for position, size, and magnitude of focus, differing less than 10 % in focal pressure and less than 1 mm in focal positioning56. Acoustic simulations are, therefore, in agreement with other simulation models. Thermal predictions are more difficult and have not been experimentally validated since MR thermometry cannot be performed in the skull, and the temperature changes are expected to be very small, which creates a source of uncertainty. The thermal modeling used in BabelBrain is consistent between papers, solving the Bioheat Transfer Equation (BHTE) to simulate the heating of the skull through modeling thermal diffusion, effects of perfusion, and acoustic power36,51. Estimates of thermal rise from models of attenuation and absorption are used to predict the effects in the skull with TUS based on other published literature57, serving as the best alternative to date for simulating skull heating. As seen in the methods described, the user specifies the duration of ultrasound, time off after ultrasound, duty cycle, and desired spatial-peak pulse-average intensity in brain conditions, of which the safety parameters are calculated as a function of parameter selection. The user must, therefore, inspect all parameters to be employed to ensure they are within safety requirements. Various neuronavigational systems used to position the transducer over the region of interest have also been seen to have a target registration error of approximately 3 mm38,58,59. However, depending on the frequency employed and the size of the target, the focal spot may still be large enough to cover the target, even with 3 mm of error. Although this may be a limitation, it still offers a greater benefit over alternative navigational systems due to the invasiveness of stereotactic localization methods.
BabelBrain streamlines planning steps required to complete TUS experiments in humans as it is efficient, easy to use, and is the only open-source accessible planning tool to date. This allows for transparency in experimental methods in published literature, as the scripts used are openly accessible. The method described completes data preparation and image processing steps, eliminating the need for customized scripts or any external collaboration to run simulations. It, therefore, removes a major barrier for experimenters, allowing them to complete planning and simulations directly and enabling more laboratories to complete successful TUS experimentation. In addition to prospective planning of experiments, this method also has potential applications in planning cavitation-mediated FUS therapy36, such as the opening of the blood-brain barrier59. Overall, this method provides an effective, streamlined approach for planning and implementing TUS experimentation in humans and the development of TUS for use in clinical settings.
The authors have nothing to disclose.
This work was supported in part by a Natural Sciences and Engineering Research Council of Canada Discovery Grant, the INNOVAIT program, Cumming Medical Research Fund, Canada Foundation for Innovation (Project 36703), Hotchkiss Brain Institute CAPRI Grant, and Parkinson Association of Alberta Funding. GBP acknowledges support from the Canadian Institutes for Health Research (FDN-143290) and the Campus Alberta Innovates Chair Program.
128-channel amplifier unit | Image Guided Therapy | This unit drives the H-317 transducer | |
24-channel head coil | General Electric | ||
3D printer | Raise3D | Pro2 | Filament thickness of 1.75mm. |
3T MRI scanner | General Electric | Discovery 750 HD | MR Console version DV26.0_R05_2008 |
BabelBrain | Samuel Pichardo (University of Calgary) | Version 0.3.0 | Accessible at https://github.com/ProteusMRIgHIFU/BabelBrain. Executes thermal and acoustic simulations. |
Blender | Blender Foundation | Version 3.4.1 | Accessible at https://www.blender.org. Blender is called automatically by BabelBrain. |
Brainsight | Rogue Research | Version 2.5.2 | Used for target identification, trajectory planning, and execution of TUS delivery sessions. |
Chair and chin/head holder | Rogue Research | To be used during TUS delivery session to ensure stability of participant’s head for optimized targeting. | |
Custom-made coupling cone | University of Calgary team | 3D printed cone in acrylonitrile butadiene styrene (ABS), only required for H-317 transducer. | |
dcm2niix | Chris Rorden (University of South Carolina) | Version 1.0.20220720 | Accessible at https://github.com/rordenlab/dcm2niix/releases. Used for pre-processing subject MR images. |
Fiducials and headband or glasses | Brainsight, Rogue Research | ST-1325 (subject tracker), LCT-583 (large coil tracker) | Headband or glasses can be interchangeably used. |
Headphones | Beats | Fit Pro True Wireless Earbuds | Wireless Bluetooth earbuds with disposable tips. |
MacBookPro | Apple | M2 Max, 16”, 64GB RAM | Computer for completing trajectory planning and simulations |
SimNIBS | Axel Thielscher (Technical University of Denmark) | Version 4.0.0 | Accessible at https://simnibs.github.io/simnibs/build/html.index.html |
Syringe(s) | 10 mL, 60 mL | Used to add additional ultrasound gel to fill air pockets. | |
Transducer | Sonicconcepts | H-317 | Other supported transducers include CTX_500 (NeuroFUS, Sonicconcepts), Single element, H-246 (Sonicconcepts), and Bsonix (Brainsonix) |
Transducer film | Sonicconcepts | Polyurethane membrane | Interface between transducer and the subject |
Ultrasound gel | Wavelength | Clear Ultrasound Gel | Coupling medium. |
Windows Laptop | Acer | Aspire A717-71G, Intel Core i7-7700HQ, 16 GB RAM | System used to control 128-channel amplifier and generate sound through the headphones |
.