High-throughput ultrasound neuromodulation in awake and freely behaving rats

Abstract

Transcranial ultrasound neuromodulation is a promising potential therapeutic tool for the noninvasive treatment of neuropsychiatric disorders. However, the expansive parameter space and difficulties in controlling for peripheral auditory effects make it challenging to identify ultrasound sequences and brain targets that may provide therapeutic efficacy. Careful preclinical investigations in clinically relevant behavioral models are critically needed to identify suitable brain targets and acoustic parameters. However, there is a lack of ultrasound devices allowing for multi-target experimental investigations in awake and unrestrained rodents. We developed a miniaturized 64-element ultrasound array that enables neurointerventional investigations with within-trial active control targets in freely behaving rats. We first characterized the acoustic field with measurements in free water and with transcranial propagation. We then confirmed in vivo that the array can target multiple brain regions via electronic steering, and verified that wearing the device does not cause significant impairments to animal motility. Finally, we demonstrated the performance of our system in a high-throughput neuromodulation experiment, where we found that ultrasound stimulation of the rat central medial thalamus, but not an active control target, promotes arousal and increases locomotor activity.

1. Introduction

Low-intensity focused ultrasound is rapidly emerging as a promising and versatile neurointerventional tool [1,2]. Its ability to modulate neural activity transiently and reversibly in deep brain structures, with high spatiotemporal specificity, safely, and noninvasively constitutes a highly attractive feature that complements current clinical neuromodulation modalities. Effective direct ultrasound neuromodulation has been demonstrated in humans [3–6] as well as in large [7–13] and small [14–17] animal models. However, its underlying neurobiological mechanisms have remained elusive, and it is unclear how different parameters including pulsing frequency, duty cycle, stimulus duration, and acoustic intensity contribute to the elicited effects [2,18,19]. Critically, the neuromodulatory action of ultrasound has been linked to peripheral auditory responses [20,21], highlighting the need for carefully controlled investigations to disentangle direct neural effects from nonspecific auditory or other sensory confounds [22], particularly in rodent models.

Ideally, such investigations would 1) include active control targets within the same trial where the auditory effects are kept consistent [23]; 2) test multiple pulsing parameters with a randomized design to avoid order effects; and 3) measure behavioral readouts in clinically relevant models. However, there is a lack of ultrasound systems allowing for such flexible targeting in freely behaving rodents. Previous wearable systems require mechanical positioning for beam steering, effectively preventing the use of within-trial active controls [14,24,25]. In turn, array devices that can electronically steer the ultrasound beam without mechanical targeting generally necessitate anesthesia or restraint due to the bulkiness of the system [26]. Here we fill this gap by developing a rat-wearable ultrasound device with electronic steering capabilities for high-throughput, multi-target neurointerventional investigations. Our device contains a 2-D array of ultrasound emitters in a miniaturized form factor (Fig. 1A–B) and can target any brain region in the behaving rat, enabling rapid testing of multiple targets and acoustic parameters. We demonstrate the potential of our device by performing ultrasound stimulation of the central medial thalamus (CMT), a brain region that plays a key role in arousal regulation, with a randomized study design including an active control target.

Fig. 1. Miniaturized ultrasound array for whole-brain targeted neurointervention.

(A) Schematic representation of the rat-wearable device and ultrasound focusing (left) (not to scale). The 64 ultrasound transducer elements are randomly distributed in a hemispherical surface (radius of curvature of 15 mm) to minimize side-lobe artifacts in the emitted ultrasound field (right). (B) Picture of the manufactured device and 3-D printed clip-on holder and comparison with a quarter US Dollar coin. (C) Acoustic field for a 1-MHz planar array simulated in k-Wave [27] with different aperture sampling approaches. A secondary grating lobe is visible with regular under-sampling (1.2-λ pitch) due to spatial aliasing. The amplitude of the grating lobe is > 60% the peak negative pressure in the main lobe. The random element distribution eliminates this grating lobe artifact and diffuses the aliasing noise over a larger volume. (D) Power transmission coefficient. (E) Magnitude (top) and phase (bottom) of the electrical input impedance of the 64 transducer elements. The plots display the individual elements and the mean magnitude and phase (thick line). (F) Representative emitted 40-cycle sinusoidal pulse measured in water with 0.23 MPa peak-to-peak pressure. (G) Output pressure characterized as a function of the Verasonics transmit power controller (TPC) voltage. The peak negative pressure was measured in water at the geometric focus. Data are mean ± s.d. from n = 5 independent sessions. The red dashed line displays the linear fit.

2. Methods

2.1. Ultrasound simulations

Ultrasound simulations were performed in MATLAB (MathWorks, Inc., Natick, MA) using the k-Wave simulation toolbox version 1.4 [27]. For the plots in Fig. 1C, we simulated a 1-MHz planar array with different aperture sampling approaches and using a homogeneous attenuating medium with a speed of sound of 1540 m/s and a density of 1000 kg/m³. The acoustic beam was focused at (0, 6) mm. For the fully populated array, we used 484 emitting elements regularly distributed with a pitch of 0.75 mm (λ/2). For the regular and random under-sampling simulations, we used 81 ultrasound emitters that were either positioned in a regular grid with a pitch of 1.2λ or randomly distributed. The lateral width of the aperture was kept constant between the three cases.

We then created a synthetic model of the wearable array by using the exact coordinates and orientation of each transducer element. We calibrated the input pressure distribution (at t₀) with free-water simulations to obtain peak pressures at the geometric focus that matched those from the hydrophone measurements. These input values were used to estimate the pressure and intensity at the target (Table 1) and in all the following acoustic and thermal simulations. We used a CT scan of a rat head that was segmented in ImageJ into bone, brain, and water compartments. We identified anatomical landmarks (bregma and lambda) in the CT images to determine the exact position of the array with respect to the rat skull. The segmented CT images were registered to match the array orientation and were interpolated in the simulation space. We used target coordinates from the in vivo experiments to assess the intensity and temperature distribution at the target and in the skull.

Table 1.

Calculation of pressures and intensities of ultrasound neuromodulation stimuli.

TPC Voltage (V)	2	4	6	8	10	12
Measured peak pressure at geometric focus in watera (MPa)	0.24	0.56	0.88	1.2	1.51	1.83
Simulated peak pressure at CMTb (MPa)	0.13	0.3	0.47	0.63	0.8	0.97
ISPPAc (W/cm2)	0.52	2.83	6.99	13	20.58	30.22
ISPTAd (W/cm2)	0.04	0.23	0.55	1.04	1.65	2.43

^aFrom hydrophone measurements.

^bSimulated at the CMT target coordinates with speed of sound and density from a segmented rat head CT scan. We calibrated the input pressure distribution (at t₀) to provide free-water peak pressure at the geometric focus that matched the hydrophone measurements.

^cThe I_SPPA was calculated considering an acoustic impedance Z = 1.55 10⁶.

^dThe I_SPTA was calculated by averaging over the inter-burst interval (10 s).

In the acoustic simulations, we assigned voxel-wise speed of sound, density, and attenuation values from the International Transcranial Ultrasonic Stimulation Safety and Standards (ITRUSST) consortium report [28]. For the bone, we used a speed of sound of 2550 m/s, a density of 1775 kg/m³, and an attenuation coefficient of 12 dB/cm (average of trabecular and cortical bone). For the brain tissue, we used a speed of sound of 1546 m/s, a density of 1046 kg/m³, and an attenuation coefficient of 0.6 dB/cm. All voxels that were not assigned to either bone or brain were considered as water (speed of sound of 1500 m/s; density of 1000 kg/m³; attenuation coefficient of 0 dB/cm). We used a sinusoidal pulse with a center frequency of 1 MHz and simulation times sufficient to reach a steady state.

In the thermal simulations, we assigned voxel-wise thermal conductivity and specific heat capacity [22] to the segmented CT data. For the bone compartment, we used a thermal conductivity of 0.32 W/m/°C and specific heat capacity of 1313 J/kg/°C; for the brain compartment, we used a thermal conductivity of 0.51 W/m/°C and specific heat capacity of 3630 J/kg/°C; for the water compartment, we used a thermal conductivity of 0.6 W/m/°C and specific heat capacity of 4178 J/kg/°C. We simulated 10 cycles of heating (80 ms) and cooling (400 ms) at the peak pressure in each voxel.

2.2. Impedance measurement

We used a 4194A Impedance/Gain-Phase Analyzer (Hewlett-Packard, Palo Alto, CA, USA) to measure the electrical input impedance of the transducer elements. During these measurements, the array face was in room-temperature water with no reflectors.

2.3. Clip-on holder design and 3-D printing

The implantable clip-on holder was designed in OnShape (OnShape, Boston, MA, USA). The implanted baseplate and clip-on holder were 3-D printed in VeroClear rigid material using a J735 printer (Stratasys, Rehovot, Israel) with Polyjet technology and with a resolution of 27 μm. Design files for the 3-D printed parts can be downloaded at https://github.com/todiian/wearable-array.

2.4. Acoustic characterization

To characterize the acoustic field of our ultrasonic array and validate the focusing scripts, we measured the emitted pressure with a lipstick hydrophone (HGL-0400, Onda Corporation, Sunnyvale, CA, USA) in degassed water using a 3-axis positioning system (AIMS III, Onda Corporation). The hydrophone was first carefully positioned at the peak-pressure location via a sequence of successive 1-D and 2-D scans. Then, the emitted peak-negative pressure was characterized at the geometric focus as a function of the Verasonics transmit power controller (TPC) voltage (Fig. 1F). The TPC voltage was set between 1.6 and 6.8 V in steps of 0.2 V. The maximum voltage was determined by the saturation point of the hydrophone preamplifier (corresponding to approximately 1 MPa). The pulse length was adjusted to ensure that a steady state was reached by the transducer while avoiding interfering reflections between the transducer and hydrophone. This measurement was repeated in five independent sessions to average out measurement noise and uncertainties in the hydrophone positioning angle with respect to the ultrasound array axis. The acoustic field was also imaged in the lateral-axial and lateral-elevation planes while varying the steering angle of the ultrasound beam. Additionally, we imaged the acoustic field through an excised skull (320 g, 10 weeks old) to assess whether the focusing and steering performance was maintained through transcranial ultrasound propagation (Fig. 2D).

Fig. 2. Pressure field characterization with free-water and transcranial propagation.

(A) Peak-negative pressure measured in free water in the lateral-axial plane with steering at 0 mm, 2 mm, 4 mm, and 6 mm in the lateral direction. The axes were centered at the peak-pressure position. (B) Peak-negative pressure measured in free water in the lateral-elevation plane with steering at 0 mm, 2 mm, 4 mm, and 6 mm in the lateral direction. The plane was placed at the peak position in the axial direction (0 mm in panel A; 0 mm in each axis indicates the focus center without steering). Aperture sub-sampling artifacts appear in the form of a distributed noise rather than well-formed grating lobes owing to the random element distribution. (C) Beam profiles in the lateral and axial directions. The profiles were sampled along the beam axis (axial) and along a perpendicular segment placed at the peak pressure location (lateral) as shown in Supplementary Fig. 1. (D) Hydrophone measurement setup with excised skull flap attached to the clip-on holder. The rostral end of the skull is on the right side of the picture. AP: anteroposterior. ML: mediolateral. DV: dorsoventral. (E) Peak-negative pressure measured in the coronal and sagittal planes. The axes were centered at the peak-pressure location. (F) Peak-negative pressure measured in the axial plane with steering at 0 mm and 2 mm in the lateral direction. The plane was placed at the peak location in the coronal plane. (G) Beam profiles in the mediolateral and dorsoventral directions without steering.

2.5. Animals

The experimental protocol for the animal procedures was approved by the Institutional Animal Care and Use Committee at Stanford University. Male and female Long Evans rats were used in all the experiments (Charles River Laboratories, Wilmington, MA, USA). All animals were 9–10 weeks old when they entered the study and weighed 354 ± 59 g (mean ± s.d.). Rats had ad libitum access to water and food for the entire duration of the experiments, were housed in a temperature-controlled vivarium on a 12-h light-dark cycle (lights on at 7 a.m.; lights off at 7 p.m.), and were acclimated to their home cage for one week before experimentation.

2.6. Surgical implant

The holder baseplate was surgically implanted and the coordinate system was fixed using a stereotaxic frame (Stoelting Co., Wood Dale, IL, USA) and the Paxinos brain atlas coordinate system [29]. Rats were anesthetized with 3.5% isoflurane in 100% oxygen, and anesthesia was maintained with 1.5% isoflurane. Body temperature was maintained at 37 °C by a warming pad with rectal probe monitoring (RightTemp Jr.; Kent Scientific, Torrington, CT, USA), and heart rate and arterial oxygen saturation were monitored using a pulse oximeter (MouseStat Jr.; Kent Scientific). The skin near the incision region was shaved using a depilatory cream and was disinfected by applying alternating povidone-iodine and 75% EtOH. After an incision was performed along the midline, the bone was cleaned with 75% EtOH and was pretreated with a bonding agent (iBOND Total Etch; Kulzer, Hanau, Germany). Three self-tapping stainless-steel screws (F2CE250; J.I. Morris Co., Oxford, MA, USA) were implanted at AP +11.3 mm, ML 0 mm and AP −12 mm, ML ± 2.9 mm via burr hole and sealed with dental cement (Tetric EvoFlow; Ivoclar Vivadent, Schaan, Liechtenstein). The upper surface of the implanted plate was positioned at DV −2.5/−2.9 mm. Special care was paid to ensure that no dental cement was placed in the propagation path to avoid distortions of the ultrasound beam. The skin was then sutured along the midline. A dose of 0.5 mg/kg Buprenex SR was administered subcutaneously for analgesia. Rats were singly housed following the surgery and were allowed to recover for 1 week before the first recording session.

2.7. Blood-brain barrier opening

In one male rat (314 g, 9 weeks old), we implanted the clip-on plate using nylon screws for MR compatibility (SPE-M1.6-6-MC; NBK America, King of Prussia, PA, USA). The rat received a bolus injection of ultrasound contrast agent (DEFINITY; Lantheus Medical Imaging, North Billerica, MA, USA; 20 μL/kg, intravenously [i.v.]), and ultrasound bursts were applied for 2 min with an estimated in-situ pressure of 0.6 MPa, 1-Hz pulse repetition frequency, and 5% duty cycle. The acoustic pressure at the target was corrected to compensate for skull insertion losses [30]. The ultrasound beam was targeted in the striatum at AP −0.8 mm, ML ± 3 mm, DV 3.6 mm. Within 1 h from sonication, rats were placed in a 7T preclinical MRI scanner (Biospec, Bruker) and were administered contrast agent (gadopentetate dimeglumine; 0.25 mL/kg; i. v.). A post-contrast T1-weighted MRI acquisition was performed to confirm blood-brain barrier permeabilization at the target.

2.8. Locomotor activity

All behavioral tests were performed during the light phase (7 a.m. – 7 p.m.) in an environmentally controlled room. Open-field locomotor activity was recorded in a custom-built apparatus consisting of a clear-floor chamber (300 mm × 400 mm × 300 mm; W × L × H) and a counterbalancing system to relieve the weight of the ultrasound array. Videos were collected using a USB camera (Logitech) placed under the chamber floor at the center of the field. Prior to the behavioral tests, rats were briefly anesthetized (<2 min) and the headgear was locked on the implanted baseplate. Centrifuged ultrasound gel was used for acoustic coupling. The animals were then placed at the center of the arena, and the video recording was started. The apparatus was thoroughly cleaned with Virkon between recording sessions to remove scent-related confounds. White noise (60 dB) was played during the sessions to attenuate any external noise. All videos were analyzed in ToxTrac [31] to track the instantaneous animal center position and quantify the distance traveled and speed. The speed was calculated by integrating the distance traveled over 1-s bins. In the spontaneous locomotion experiments, tracking started immediately upon recovery from anesthesia to evaluate animal motility during the hyperactive phase. In the CMT stimulation experiments, we allowed for a 30-min period before the start of the sonication and tracking to allow for full isoflurane clearance.

2.9. Ultrasound neuromodulation

In the behavioral testing apparatus described above, we delivered ultrasound stimuli to the CMT (AP −3.1 mm; ML 0 mm; DV 6 mm) or an active control target (AP 2.4 mm; ML 0 mm; DV 6 mm). The active control target was placed in a location symmetric to the CMT in the AP direction with respect to the center of the array (AP −0.35 mm) to maintain equivalent steering conditions and comparable pressure distribution in the skull in the two cases. The active control target location corresponded with the ventral end of the dorsal peduncular cortex. Ultrasound stimuli were made of either 1 or 3 bursts. Each burst had a duration of 5 s and was made of sinusoidal pulses with a duration of 80 ms and a pulse repetition period of 480 ms [14]. We used an inter-burst interval of 10 s, giving a duty cycle of 17%. We did not use an amplitude ramp to smoothen the envelope of the ultrasound pulses.

Within the same test session, ultrasound stimuli were delivered to the CMT or active control and with TPC excitation voltage between 2 and 12 V in steps of 2 V. We randomized the sequence of excitation voltage and target in individual stimuli in each session to avoid order effects. One stimulus was delivered every 2 min while recording continuously for subsequent locomotor activity quantification.

To calculate the values of pressure and intensity corresponding to the different excitation voltages, we first calibrated the simulation to the free-water hydrophone measurements, then we simulated the transcranial acoustic field targeted at the CMT coordinates and extracted the pressure and intensity at the target. The estimated pressure at the CMT was between 0.13 and 0.97 MPa, corresponding to a spatial peak-pulse average intensity (I_SPPA) between 0.52 and 30.22 W/cm², and a spatial peak-temporal average intensity (I_SPTA) between 0.02 and 2.42 W/cm² (Table 1).

The distance and speed time courses were quantified from the recorded videos and were time locked to the stimulus trigger. The maximum speed and cumulative distance traveled were calculated in the 0–15 s and 15–30 s time intervals for subsequent statistical analysis. These measures were normalized to the group-level mean of the 2 V stimulation, which was considered as a baseline. The 2 V (0.52 W/cm²) data points were plotted for reference but were excluded from the statistical analysis.

210. General statistical analysis

All statistical analyses were performed using custom scripts in R Studio. Statistical tests, sample sizes N, and P values are reported for each analysis in the text and figure captions.

3. Results

3.1. Transducer array design and manufacturing

We designed a hemispherical array of 64 randomly distributed ultrasonic emitters, each with a diameter of 1.3 mm (Fig. 1A). The device was designed in collaboration with and manufactured by Sonic Concepts, Inc., Bothell, WA, USA (Fig. 1B) and connects to a Verasonics research scanner for flexible and versatile sequence programming. Focusing scripts can be accessed at https://github.com/todiian/wearable-array.

The random element distribution minimizes the number of channels, and hence the complexity, bulkiness, and cost of the system, while limiting the occurrence of grating-lobe artifacts in the emitted ultrasound field (Fig. 1C). Indeed, with the randomized under-sampling approach, spatial aliasing appears as a diffused noise distributed over a larger volume rather than a well-defined grating lobe.

The manufactured array has a radius of curvature of 15 mm and an aperture diameter of 25 mm. The geometric focal depth is 8.18 mm from the edge of the housing. The RF shielded transducer housing is made of anodized aluminum for corrosion protection and is coated with a thin film of epoxy that makes it water-tight up to the cable level. The housing integrates a series of markers for unequivocal positioning when paired with the clip-on holder (Fig. 1B). The outer diameter of the housing is 29 mm. Impedance matching was achieved by means of 64 RF matching networks supplied inside an RF shielded Cannon DL-260 connector shell. The array has a center frequency of 0.95 MHz and a relative bandwidth of 19 % (0.86–1.05 MHz; Fig. 1D–G). In all the experiments, the array was operated at 1 MHz.

We designed an implantable clip-on mounting device that allows for rapid fastening of the ultrasound array on the rat’s head. This device consists of a baseplate that is chronically implanted via stereotactic surgery using three anchoring screws, and a clip-on holder attached to the ultrasound array (Fig. 1B). The holder has two lateral supporting wings to ensure a snug fit and is quickly fastened and unfastened on to the plate by a 10-deg rotation. The wearable ultrasound headgear allows the animals to move freely without severely affecting turning and rearing behaviors (Fig. 3A and Supplementary Movie).

Fig. 3. Targeting validation and headgear effects on spontaneous locomotion.

(A) Picture of a rat wearing the ultrasound headgear. The device does not impede movement even when climbing the chamber walls. (B) Annotated gadolinium-enhanced MR images of the rat brain after blood-brain barrier opening targeted at ML ± 3 mm. (C, D) Distance traveled and speed measured in a spontaneous locomotion behavioral recording after recovery from anesthesia. No significant differences in locomotion were found when wearing the ultrasound headgear (cumulative distance: P > 0.58, Headgear vs Surgery; P > 0.26, Headgear vs Baseline; max speed: P > 0.51, Headgear vs Surgery; P > 0.11, Headgear vs Baseline; two-tailed paired t-test; N = 4 rats). (E) Representative traces of body movement from animals in C-D.

3.2. Acoustic characterization

The acoustic beam can be electronically steered up to at least ±6 mm in either direction without introducing distinct grating-lobe artifacts in the pressure field, confirming the benefits of the random element distribution (Fig. 2A–C). This steering range is sufficient to cover most of the rat brain without mechanically moving the ultrasound device. The presence of beam distortions can be noted in Fig. 2B at higher steering angles due to spatial aliasing, as expected given the significant under-sampling of the ultrasound emitting aperture (pitch >1.8 mm with a wavelength of ~1.5 mm). The full-width at half maximum (FWHM) is 1.69 mm laterally and 9.28 mm axially when measured at the geometric focus (Fig. 2C). With electronic steering, the average FWHM is 2.09 ± 0.37 mm (mean ± s.d.) laterally and 9.64 ± 1.65 mm axially.

We measured the acoustic field in the presence of an excised rat skull to assess the focusing and steering performance with transcranial acoustic propagation (Fig. 2D–G). The array’s focusing (Fig. 2E) and steering (Fig. 2F) performance in this setting is comparable to the propagation in free water, and no significant distortions were introduced by the bone. The lateral FWHM is 1.78 mm, an increase of 5% compared to the propagation in free water. In the axial direction, we measured the half-width at half maximum (HWHM) as it was not possible to measure the entire beam near the skull. The axial HWHM was 4.97 mm (corresponding to an estimated FWHM of ~9.94 mm). It is worth noting that our transcranial measurements do not account for standing waves that could form in an intact skull with long ultrasound pulses.

3.3. Targeting validation and effects on spontaneous locomotion

To confirm the targeting in vivo, we performed ultrasound-mediated blood-brain barrier opening (BBBO) and assessed the BBBO targeting fidelity with contrast-enhanced magnetic resonance imaging (MRI). With the animal wearing the ultrasound headgear, we infused Definity microbubbles while electronically steering the beam to deliver ultrasound in two targets positioned at ± 3 mm laterally. The post-BBBO MRI confirmed the presence of gadolinium extravasation in the proximity of the ultrasound targets (0.32 mm mean error; Fig. 3B). The accuracy of the ultrasound targeting was comparable to that of the stereotactic positioning (0.5 mm) [29].

We then assessed the effects of wearing the ultrasound headgear on spontaneous locomotion. We designed an experimental apparatus consisting of a test chamber and a counterbalancing system to alleviate the weight of the ultrasound device and connecting cable from the rat’s head. The chamber has a transparent floor to allow for video recording from below (Supplementary Movie). In a 20-min session, we measured the distance traveled and the locomotion speed at 1) baseline; 2) after surgical implantation of the clip-on holder; and 3) while wearing the ultrasound headgear (Fig. 3C–E). We found no statistically significant differences between the three conditions, and particularly between the headgear-wearing and post-surgery recordings (P > 0.58 for cumulative distance and P > 0.51 for maximum speed; two-tailed paired t-test). Taken together, these results confirm that our wearable device can target multiple brain regions via electronic steering of the ultrasound beam and without introducing significant impediments to the animal motility.

3.4. Ultrasonic stimulation of the central medial thalamus, but not an active control target, increases locomotor activity

To demonstrate the potential of our device for high-throughput testing of brain targets and acoustic parameters in an ultrasound neuromodulation application, we implemented an ultrasonic CMT stimulation assay in freely behaving rats. The central thalamus is a deep-brain structure that plays a key role in regulating arousal and awareness. Electrical and optogenetic stimulation of this region acutely promotes transitions toward an active and aroused state, with resulting increases in exploratory and goal-directed behavior [32–34]. Therefore, we hypothesized that ultrasound-mediated stimulation of the CMT would facilitate an aroused state, and we used locomotor activity as a robust and validated behavioral readout to indirectly measure arousal.

During the animal’s inactive phase (7AM - 7PM) and within the same session, we targeted the CMT or an active control target (Fig. 4C). While wearing the headgear and in the test chamber, rats received ultrasound stimuli every 2 min directed at either the CMT or active target and with intensities (I_SPPA) between 0.52 and 30.22 W/cm² (Table 1). The stimulation conditions were shuffled within each session. We delivered 80-ms pulses with a 480-ms inter-pulse interval and a burst duration of 5 s (Fig. 4B). We used either a single-burst stimulus or a sequence of 3 bursts with a 10-s inter-burst interval.

Fig. 4. Ultrasound modulation of the central medial thalamus increases locomotion.

(A) Schematic representation of the high-throughput ultrasound stimulation setup with behavioral locomotion readout. (B) Representation of the 3-burst ultrasound stimulation sequence. BD, burst duration; IBI, inter-burst interval; PD, pulse duration; PRF, pulse repetition frequency. (C) Intensity beam plot (heatmap) simulated in k-Wave using a synthetic model of the wearable array and a segmented CT scan of a rat head. The two beams were targeted to central medial thalamus (CMT) and to an active control positioned symmetrically in the anteroposterior direction. Top plots display a sagittal view. Bottom plots display an axial view. Scale bar: 1 mm. (D) Normalized distance traveled and maximum speed quantified from the behavioral readout (15–30 s interval) in response to ultrasound stimulation of the CMT or active control target. *P < 0.05, **P < 0.01, two-tailed paired t-test between voltage values in each group. ^#P < 0.05, ^##P < 0.01, two-tailed paired t-test between CMT and active control. N = 39 trials in 6 rats for 1-burst stimulation. N = 21 trials in 3 rats for 3-burst stimulation. (E) Speed time series for 1- and 3-burst stimulation with I_SPPA of 20.58 W/cm². The blue lines represent the ultrasound stimuli. Data presented as mean ± s. e.m.

We then quantified the instantaneous distance traveled and locomotion speed from the recorded videos in the 0–15 s and 15–30 s time intervals. In the distance traveled in the 15–30 s interval (Fig. 4D), a three-way mixed-effects ANOVA with between-subjects factor of burst length (1 or 3) and within-subjects factors of target (CMT or active control) and intensity showed a significant effect of target (F_1,58 = 18.72, P = 6.04E-05) and significant burst length × target (F_1,58 = 5.8, P = 0.019), intensity × target (F_4,232 = 3.06, P = 0.018), and burst length × intensity × target (F_4,232 = 2.93, P = 0.022) interactions. Similarly, a three-way mixed-effects ANOVA performed on the maximum speed (Fig. 4D–E) showed significant effects of target (F_1,58 = 12.84, P = 6.95E-04) and significant burst length × target (F_1,58 = 7.73, P = 0.007) and intensity × target (F_4,232 = 2.77, P = 0.028) interactions. No significant effects were found in the 0–15 s time interval. To further assess the ultrasound CMT stimulation effects on locomotor behavior with 1 and 3-burst stimulation separately (15–30 s interval), we performed two-way ANOVAs stratified by burst length with within-subjects factors of target and intensity. In the 3-burst stimulation, there was a significant effect of target (F_1,20 = 14.77, P = 0.001) and significant intensity × target interaction (F_4,80 = 3.53, P = 0.011) in the distance traveled, and significant effects of target (F_1,20 = 15.56, P = 8E-04) and intensity × target interaction (F_4,80 = 2.9, P = 0.027) in the maximum speed. Posthoc pairwise comparisons (two-tailed paired t-test) showed significant differences between the CMT and active control target in both measures with 3-burst stimulation and I_SPPA between 13 and 30.22 W/cm². No significant differences were found with 1-burst stimuli.

To estimate the maximum temperature rise at each voxel in the simulation grid with the I_SPPA level that was maximally effective in the CMT neuromodulation experiment (20.58 W/cm²), we performed temperature simulations that mirrored the sonication sequence (80 ms on/400 ms off) for 10 cycles (1 burst). We estimated a maximum temperature increase of 0.31 °C at the target and 1.63 °C in the skull at the end of the last 80 ms on cycle (Fig. 5A).

Fig. 5. Thermal simulations and pressure distribution at the skull.

(A) Temperature rise (heatmap) for the central medial thalamus (CMT) target simulated in k-Wave using a synthetic model of the wearable array and a segmented CT scan of a rat head. The plots display a sagittal and an axial view. (B) Pressure distribution within the skull when targeting the CMT or active control and with an unfocused emission. AP: anteroposterior. ML: mediolateral. Scale bars: 1 mm.

To confirm that the pressure distribution in the skull was comparable between the CMT and active control targets, we extracted the simulated peak negative pressure at the skull level and averaged three axial slices (0.75 mm thickness) starting at the bregma dorsoventral position (Fig. 5B). The simulated pressure fields were overall comparable between the CMT and active control. However, the pressure was distributed differently in the two conditions and the peaks were concentrated anterior or posterior to bregma depending on target location. In addition, we simulated an unfocused emission that could be used as an alternative control condition (Fig. 5B and Supplementary Fig. 2). In this case, the pressure was distributed more randomly in the skull and the peak pressure was lower than in the focused emissions, as expected. Unfocused and active control sonications could potentially be combined in the same session for a more thorough comparison.

Taken together, our results demonstrate that focused ultrasound delivered to the CMT, but not to the active control target, increased locomotor activity in an intensity- and burst-length-dependent manner. This may indicate an acoustic-dose-dependent stimulatory effect of the CMT that mediated an overall increase in arousal. Importantly, these findings confirm the potential of our headgear device to facilitate high-throughput ultrasound neuromodulation studies with randomized experimental designs using within-trial active control targets.

4. Discussion

In this paper, we developed a miniaturized 64-element ultrasound array for neurointerventional studies in freely behaving rats. Our device enables high-throughput investigations to flexibly test large numbers of parameters and brain targets in a single session, using within-trial focused or unfocused active controls. We designed a 3-D printed clip-on mounting device to facilitate rapid fastening of the ultrasound array on the rat’s head. Ultrasound targeting is achieved by means of electronic steering, leveraging the benefits of a random element distribution to minimize the occurrence of grating lobes in the acoustic field. We integrated our device with a behavioral assay to evaluate locomotion-related measures in an open field. Notably, this system could be used in combination with any behavioral paradigm to test ultrasonic interventions in clinically relevant rat behavioral models. In addition, electrophysiological readouts could be integrated by using implanted surface or depth electrodes.

We demonstrated the potential of our wearable array for high-throughput neuromodulation investigations in an ultrasonic CMT stimulation experiment. The central thalamus is a key region of the brain network responsible for regulating arousal, wakefulness, and motivation [35]. Bilateral electrical stimulation of the central thalamus in a patient who had remained in a minimally conscious state for 6 years before the intervention produced acute changes in arousal and restored behavioral functions in a six-month double-blinded trial [32]. Behavioral and neurophysiologic transitions toward an awake and aroused state have also been demonstrated in primates and rodents by electrical and optogenetic stimulation of the central lateral and central medial thalamic nuclei [33,34,36,37]. Notably, one study investigated the dose-response relationship of the electrical stimulation in a cognitive fatigue task in nonhuman primates and found that cognitive performance changes were dependent on the amplitude of the electrical stimuli with an inverted-U function [36]. Interestingly, in our experiments we observed a similar inverted-U-shaped relationship between the behavioral responses and the total acoustic dose. Specifically, we found that behavioral measures of arousal peaked in response to ultrasound CMT stimulation with an intensity of 20.58 W/cm² and decreased at higher acoustic doses. While in this experiment we only focused on the dose-response relationship as a function of the acoustic intensity, other pulsing parameters such as the duty cycle, pulse repetition frequency, and stimulation duration may also contribute to the neuromodulatory effect [38,39]. Future work should leverage our electronically steered wearable array and the flexibility of the Verasonics platform to design pulsing sequences to map this parameter space.

One limitation of our study is that we did not perform histological analyses to determine the presence of ultrasound-related tissue damage. However, a number of prior studies have documented the safety of ultrasound neuromodulation in animals and humans, in some cases using I_SPTA beyond the FDA recommended limits of 720 mW/cm² (Ref. [40]). In one recent paper, four nonhuman primates were exposed to up to 147 sonications over a period of 2 years. Post-ultrasound MR imaging did not reveal structural damage or evidence of chronic or acute trauma compared to the pre-ultrasound MRI [41]. Another study performed histological analyses in two nonhuman primates and thirteen sheep following exposure to ultrasound with I_SPTA up to 26.5 W/cm², finding no tissue damage attributable to the ultrasound treatment [9]. In humans, no signs of ultrasound-related histological damage or apoptosis were observed in patients with refractory temporal lobe epilepsy scheduled for surgical resection and receiving sonications at the resected site with I_SPTA up to 5.76 W/cm² (Ref. [42]). We performed thermal simulations to assess the temperature rise at the target and in the skull, and we found that with I_SPPA of 20.58 W/cm² and I_SPTA of 1.65 W/cm², the estimated temperature rise was 1.6 °C in the skull and 0.3 °C at the CMT target. Importantly, in our neuromodulation experiments rats were exposed to up to 36 sonications in a single session, and some rats underwent up to 3 sessions of ultrasound neuromodulation. Notwithstanding, we did not observe any signs of behavioral deficits and all animals appeared healthy for the entire duration of the study.

Another limitation of the current work is that, due to the limited emitting aperture, the ultrasound focus presented a relatively large FWHM in the axial direction (~10 mm). As a result, in our neuromodulation experiment other brain regions were likely exposed to ultrasound in addition to the CMT. Located dorsal to the CMT, the mediodorsal thalamic nucleus has been shown to contribute to regulating spontaneous and psychostimulant-induced locomotor activity [43]. However, in these pharmacological manipulations the hyperlocomotor response was long-lasting (~60 min) compared to our experiment and presented a monotonic dose-response relationship. The medial habenula, located between the thalamus and the hippocampus, plays a role in exercise motivation and in regulating hedonic state [44]. Optogenetically stimulating the medial habenula induces a locomotor response, and it is therefore possible that this region mediated some of the responses observed in our experiments. Future work should focus on determining the extent of brain territory activated in response to the ultrasound stimuli by imaging early gene expression biomarkers, such as c-Fos, after ultrasound exposure.

Collectively, our data demonstrate the potential of our wearable ultrasound array to perform high-throughput ultrasound neuromodulation studies with within-trial active control targets. Such studies may facilitate the testing of pulsing sequences for ultrasound neuromodulation and the identification of brain targets with potential therapeutic efficacy in clinically relevant rat behavioral models.

Data and materials availability

CAD models and focusing scripts are available at https://github.com/todiian/wearable-array. Data and codes are available from the corresponding authors upon reasonable request.