LABORATORY INVESTIGATION J Neurosurg 128:875–884, 2018

Noninvasive and thalamic mapping with low-intensity focused ultrasound

Robert F. Dallapiazza, MD, PhD,1 Kelsie F. Timbie, PhD,2 Stephen Holmberg, CINM,6 Jeremy Gatesman, BA, LVT,3 M. Beatriz Lopes, MD, PhD,4 Richard J. Price, PhD,2 G. Wilson Miller, PhD,5 and W. Jeffrey Elias, MD1

Departments of 1Neurosurgery, 2Biomedical Engineering, 3Comparative Medicine, 4Pathology (Neuropathology), and 5Radiology, and 6Impulse Monitoring, University of Virginia, Charlottesville, Virginia

OBJECTIVE Ultrasound can be precisely focused through the intact human skull to target deep regions of the for stereotactic ablations. Acoustic energy at much lower intensities is capable of both exciting and inhibiting neural tissues without causing tissue heating or damage. The objective of this study was to demonstrate the effects of low-intensity focused ultrasound (LIFU) for neuromodulation and selective mapping in the thalamus of a large-brain animal. METHODS Ten Yorkshire swine (Sus scrofa domesticus) were used in this study. In the first neuromodulation experi- ment, the lemniscal sensory thalamus was stereotactically targeted with LIFU, and somatosensory evoked potentials (SSEPs) were monitored. In a second mapping experiment, the ventromedial and ventroposterolateral sensory thalamic nuclei were alternately targeted with LIFU, while both trigeminal and tibial evoked SSEPs were recorded. Temperature at the acoustic focus was assessed using MR thermography. At the end of the experiments, all tissues were assessed histologically for damage. RESULTS LIFU targeted to the ventroposterolateral thalamic suppressed SSEP amplitude to 71.6% ± 11.4% (mean ± SD) compared with baseline recordings. Second, we found a similar degree of inhibition with a high spatial res- olution (~ 2 mm) since adjacent thalamic nuclei could be selectively inhibited. The ventromedial thalamic nucleus could be inhibited without affecting the ventrolateral nucleus. During MR thermography imaging, there was no observed tissue heating during LIFU sonications and no histological evidence of tissue damage. CONCLUSIONS These results suggest that LIFU can be safely used to modulate neuronal circuits in the central ner- vous system and that noninvasive brain mapping with focused ultrasound may be feasible in humans. https://thejns.org/doi/abs/10.3171/2016.11.JNS16976 KEY WORDS neuromodulation; noninvasive; brain mapping; low intensity focused ultrasound; thalamus; somatosensory evoked potentials; functional neurosurgery

coustic energy has long been known to influence Advances in noninvasive, transcranial delivery of ultra- the activity of electrically excitable tissues, in- sound over the past 15 years have renewed an interest in cluding muscle, peripheral nerves, and the central its use for neurosurgical applications.3–6,18,19,28 Transcranial nervousA system.1,13,15,16 During the 1950s, high-intensity HIFU can be delivered in a precise, highly localized man- focused ultrasound (HIFU) was used experimentally to ner with millimeter accuracy to induce tissue ablation in reversibly inhibit neuronal activity through moderate heat- deep cerebral structures in the . The clinical ing below the threshold for tissue ablation,14 and it was effects of HIFU therapy have been highlighted in several used clinically to treat patients with movement disorders recent clinical trials involving patients with movement dis- and brain tumors.17,27 orders and psychiatric diseases.2,9,11,20,25,26

ABBREVIATIONS FUS = focused ultrasound; H & E = hematoxylin and eosin; HIFU = high-intensity focused ultrasound; ISA = spatial average intensity; LFB = Luxol fast blue; LIFU = low-intensity focused ultrasound; PRFS = proton resonance frequency shift; SSEP = somatosensory evoked potential; VPL = ventroposterolateral thalamic nucleus; VPM = ventroposteromedial thalamic nucleus. SUBMITTED April 16, 2016. ACCEPTED November 17, 2016. INCLUDE WHEN CITING Published online April 21, 2017; DOI: 10.3171/2016.11.JNS16976.

©AANS 2018, except where prohibited by US copyright law J Neurosurg Volume 128 • March 2018 875

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC R. F. Dallapiazza et al.

Perhaps the most exciting, yet largely unharnessed, Upon completion of electrophysiological recordings in potential for transcranial ultrasound is for neuromodula- all 8 animals, the contralateral (left) thalamus was target- tion and noninvasive brain mapping with low-intensity ed first with LIFU and subsequently with HIFU sonica- focused ultrasound (LIFU). The mechanisms of LIFU are tions during real-time MR thermography. Two additional nonthermal and are thought to be mediated by mechanical animals underwent HIFU ablations only, during electro- forces within the brain tissue.31,32 LIFU shares many of physiological recordings. Tissue from all 10 animals was the appealing characteristics of HIFU. It can be focused submitted for histological analysis. through the human skull to target deep cerebral structures without affecting intervening tissues, while demonstrating Anesthesia and a high spatial accuracy that is not available with current Anesthesia was induced with a single intramuscular noninvasive neuromodulation methods, such as transcra- injection of tiletamine and zolazepam (Telazol, 6 mg/kg) nial magnetic stimulation or transcranial electrical stimu- and xylazine (2 mg/kg). The animals were moved from the lation. cage area to a surgical preparation room. An intravenous Most current studies of LIFU neuromodulation have catheter was place in the marginal vein of each ear. An en- applied ultrasound to the of rodents to elicit 21,29,33,34 dotracheal tube was placed and secured to the mandible. muscle contractions. Limb, tail, whisker, and eye Anesthesia was maintained with a continuous infusion of muscles can be independently activated, depending on the propofol at 10 mg/kg/hr. Animals were placed on a venti- LIFU focus. However, somatotopic mapping is difficult lator (10 ml/kg tidal volume) at a rate of 18 breaths/minute due to the mismatch in size between the acoustic focus of room air. Prior to SSEP testing, an intravenous drip of and the rodent brain.21 More recently, Legon et al.24 and 22 rocuronium (200 mg diluted in a 500-ml bag of normal Lee et al. applied LIFU to the human somatosensory saline) was initiated. The drip rate was set to administer cortex, demonstrating that ultrasound could suppress me- 2.0–2.5 mg/kg/hr. The effects of paralytics were assessed dian nerve evoked potentials on EEG and stimulate sub- using physical indicators such as palpebral reflex, eye po- jective somatosensory phenomena. sition, toe pinch, and jaw tone. Depth of anesthesia and Despite these exciting advances, LIFU neuromodula- analgesia were monitored by continuous measurements of tion needs to be translated and refined in large-brain ani- 8,23 heart rate and oxygen saturation. mal models. Furthermore, prior research has not fully To implant the epidural electrode, the scalp was first highlighted the capabilities of LIFU neuromodulation— infiltrated with 0.25% bupivicaine (Marcaine), and then namely, its ability to noninvasively penetrate deep cortical a U-shaped incision was made to reflect the scalp posteri- structures (beyond the ) with a high spatial orly. A 4 × 4–cm craniectomy centered on the bregma was resolution (millimeter scale). In this article, we describe a performed using a high-speed drill, and the large-brain animal model that targets LIFU to the somato- was kept intact. A 4-contact epidural recording electrode sensory thalamus, alters somatosensory evoked potentials was placed over the right and tucked (SSEPs) for long durations, is selective and precise within laterally out of the ultrasonic beam path. The electrode 2 mm, and does not result in tissue heating or histological was secured to the dura and tunneled posteriorly for elec- damage. trophysiological recordings.

Methods Electrophysiology and SSEPs Animals and Experimental Design A 32-channel recording system with 9 channels for All experiments were approved by the University of stimulation was used to measure SSEPs in swine (Cadwell Virginia Animal Care and Use Committee. Female York- Cascade Pro). MR-compatible platinum electrodes (Gen- shire swine (Sus scrofa domesticus) weighing 25–30 lbs, uine Grass F-E2–24) were used for bipolar stimulation age 6–7 weeks, were used in these experiments. Ten ani- of the trigeminal (snout), median (forelimb), and tibial mals were used in total. (hindlimb) nerves. Recording electrodes were placed mid- LIFU neuromodulation was assessed in the first cohort line at the front of the skull, midline at the back of the of 4 animals. The right thalamus was targeted with LIFU. skull, 4 cm laterally in each direction from midline at the The left forelimb (median nerve) was stimulated, and re- back of the skull, over the cervical spine, and bilaterally cordings were made from an epidural grid placed over over the brachial plexus in the shoulder. A 1 × 4 cortical the right cerebral hemisphere. Control experiments were grid was implanted over the right lateral convexity of the conducted during alternate sonications; the contralateral cortex to optimize recordings. All recordings were refer- (left) thalamus was targeted with LIFU during left median enced to a cephalic recording electrode with the exception nerve stimulation and right cortical monitoring. of recordings from the brachial plexus electrodes, which Thalamic mapping with LIFU was assessed in a sec- were referenced to the contralateral brachial plexus elec- ond cohort of 4 animals. The right thalamus was alternate- trode. ly targeted for mapping with LIFU in 3 separate locations. The peripheral stimulation frequency ranged between The first location was anterior to the sensory thalamus. 2 and 3 Hz, and the stimulation amplitude varied from 12 The second location was the ventromedian thalamic nu- to 15 mA with 0.3-msec square wave pulse duration. Fifty cleus, and the third location was the ventroposterolateral to 300 trials were averaged, and stable baseline recordings thalamic nucleus. SSEPs were recorded in sequence from were obtained for 10–20 minutes prior to ultrasound expo- the snout (trigeminal nerve) and hindlimb (tibial nerve). sure. The trigeminal nerve cortical response was typically

876 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC Selective neuromodulation with LIFU observed 12–13 msec after the stimulation artifact, the ed fast spin-echo images of the brain were acquired. These median nerve cortical response was 14–16 msec after the images were used to identify the desired thalamic targets stimulation artifact, and the tibial nerve cortical response and determine their spatial coordinates. The patient table was 16–18 msec after the stimulation artifact. After LIFU was then slid all the way out of the magnet bore to prepare sonications, SSEP amplitudes were compared with aver- for the SSEP measurements of LIFU sonications. age baseline recordings. Data were expressed as a per- centage of baseline values and were compiled as mean ± Clinical 650 and 220 kHz FUS Systems standard deviation. A second set of experiments were performed in a simi- lar manner using 2 different multi-element ultrasound MR-Guided Focused Ultrasound transducers, each coupled to a 3-T MRI system (Discov- Ultrasound Pulsing Protocol ery MR750, GE Healthcare). The first of these 2 trans- All neuromodulation experiments (LIFU sonications) ducers was an Exablate Neuro 4000 (InSightec), which is 2 were performed at 25–30 W/cm (ISA [spatial average in- approved by the FDA for thalamotomy in human patients, tensity]), with a pulse duration of 43.7 msec and a pulse and can be adjusted to deliver ultrasound at frequencies repetition time of 100 msec (duty cycle 43.7%, Fig. 1c). ranging from 620 to 720 kHz. For this study, it was pro- The sonication duration was 40 seconds in total. In all grammed to operate at 710 kHz, but because it was de- studies, the animal was coupled to the transducer by a de- signed to operate at 650 kHz and is generally described gassed water bath. HIFU sonications were performed by as the 650-kHz device, we are using that designation here applying continuous-wave power at 20 W for 20 seconds as well. The other ultrasound transducer used in this set and were used to verify targeting and MR thermometry of experiments was a 220 kHz transducer (InSightec) that results. These HIFU sonications were performed only af- is not currently approved for clinical use. The 650-kHz ter all neuromodulation LIFU sonications had been com- transducer consists of a hemispherical (30-cm diameter) pleted in the animal. 1024-element phased-array transducer with approximate- ly a 4 × 4 × 6 mm focal region and 0.72-mm accuracy. The Preclinical 1.14-MHz FUS System 220-kHz transducer consists of a hemispherical (30-cm × This device (RK-100, FUS Instruments, Inc.) utilizes diameter) 990-element phased-array transducer with a 4 × a single-element spherically focused transducer powered 4 6 mm focal region and 0.72-mm accuracy. by a 50-dB amplifier and driven by a function generator (33210A, Agilent) at 1.145 MHz. The focal region is ap- MR Thermography proximately 1.5 × 1.5 × 7.5 mm. The device uses a motor- MR thermography was performed in each animal dur- controlled 3-axis positioning system to adjust the location ing separate LIFU and HIFU sonications, after all SSEP of the acoustic focus, with sub-millimeter accuracy. measurements had been completed. LIFU sonications were performed first, followed by contralateral HIFU son- Clinical 3-T MRI Scanner ications. The time between all sonications was at least 5 The preclinical 1.14-MHz FUS system was placed on minutes. Once an ablation was performed (recorded tem- the patient table of a whole-body 3-T MRI scanner (Trio, perature increase > 20°C), the experiment was considered Siemens Medical Solutions). Before positioning the ani- complete and no additional data were collected. Tempera- mal on the FUS platform, the coordinate systems of the ture was monitored in a single slice through the ultrasound FUS and MRI systems were synchronized by using MR focus, by acquiring a time series of temperature-sensitive thermometry to locate a HIFU-induced focal temperature gradient-echo MR images beginning before the start and rise in an anechoic Zerdine phantom (CIRS, Inc). The ani- continuing beyond the end of each sonication period (MR mal was then positioned supine on the FUS platform with parameters: slice thickness 3 mm, flip angle 25°, TR 39 its brain directly above the upward-facing transducer. The msec, TE 10 msec, FOV 250 mm, matrix 256 × 256, read- patient table was slid into the magnet bore, and T2-weight- out bandwidth 80 Hz/pixel, acquisition time 5 sec/image).

FIG. 1. Ultrasound parameters and characterization. The single-element 1.14-MHz focused transducer has a narrow 1 × 1 mm focus in the lateral dimension (a) and a 5-mm focus in the axial direction (b), with a focal length of 5.6 cm. Neuromodulation treat- 2 ments were performed with a 43.7% duty cycle over 40 seconds (c), with an ISA of 25–30 W/cm maintained across all 3 ultrasound systems used (1.14 MHz, 650 kHz, and 220 kHz). MI = mechanical index; ms = msec; s = seconds. Figure is available in color online only.

J Neurosurg Volume 128 • March 2018 877

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC R. F. Dallapiazza et al.

Maps of temperature change relative to the pre-sonication TABLE 1. Parameters used in the KZK equation from the baseline were calculated from the resulting phase im- frequency domain ages using the standard proton resonance frequency shift Material & Parameter Value Units (PRFS) method,7,19 and maps of absolute temperature ver- sus time were constructed assuming a baseline body tem- Water perature of 39°C. Speed of sound 1482 m/sec Mass density 1000 kg/m3 Theoretical Computational Temperature Increases With Absorption at 1 MHz 0.217 dB/m LIFU and HIFU Exponent of absorption vs frequency curve 2 — The theoretical temperature rise was calculated using Nonlinear parameter 3.5 — the HIFU Simulator V1.2, which first integrates the axi- Material transition distance 3.1 cm symmetric Khokhlov–Zabolotskaya–Kuznetzov (KZK) equation from the frequency-domain and then inputs the Gray matter temporal average intensity and heating rate into the bio- Speed of sound 1550 m/sec heat transfer (BHT) equation to determine the temperature Mass density 1045 kg/m3 rise at the focus and the total thermal dose delivered. The Absorption at 1 MHz 80 dB/m parameters used in the simulations are provided in Tables Exponent of absorption vs frequency curve 1.35 — 1 and 2. Sonication parameters and beam path geometry Nonlinear parameter 6.9 — were determined from experimental settings, and tissue properties for human gray matter were collected from pub- Transducer lished data (Tables 1 and 2). Resolution was adjusted to 1 Outer radius 3.5 cm mm to match the resolution of MR thermography images. Inner radius 0 cm Focusing depth 5.6 cm Histology Frequency 1.14 MHz Animals were killed with a lethal overdose of barbitu- Power 0.25 W rates, and whole were immediately dissected and placed in 10% formalin. Brains were fixed for at least 14 KZK = Khokhlov–Zabolotskaya–Kuznetzov. days prior to sectioning. Once fixation was complete, the brains were sectioned in ~ 3-mm blocks in the coronal plane. The ventrolateral thalamic nucleus was identified compared with baseline recordings (4 animals, total of 9 and removed en bloc and placed in cassettes. Tissues were sonications, Figs. 2c and d). Contralateral control sonica- embedded in paraffin wax and cut in 5-μm sections. Se- tions in the left VPL had little effect on left median nerve rial sections were cut every 100 μm and stained with he- SSEP amplitude (96.3% ± 7.3% of baseline values at 5 matoxylin and eosin (H & E) or Luxol fast blue (LFB). minutes after sonication; 4 animals, 6 sonications; Fig. 1c Gross and microscopic evaluations were performed by a and d). Peak electrophysiological suppression was seen 5 certified neuropathologist (M.B.L.) who was aware of the minutes after acoustic exposure, and the values returned experimental paradigm. Results TABLE 2. Parameters used for bioheat transfer equation to determine the temperature rise at the focus and the total thermal Thalamic Neuromodulation With LIFU dose delivered We first sought to determine whether LIFU could tem- porarily suppress the activity of thalamocortical relay neu- Material & Parameter Value Units rons in the ventroposterolateral thalamic nucleus (VPL). Water In 4 anesthetized female swine (age 6–7 weeks, weight Heat capacity 4180 J/kg/K 25–30 lbs), SSEPs were obtained by stimulating the left median nerve and recording the evoked cortical responses Thermal conductivity 0.6 W/m/K with an epidural electrode implanted over the right lateral Perfusion rate 0 kg/m3/sec convexity of the cortex. The thalamus was imaged with Gray matter 3-T MRI, and the VPL was located with the aid of a ste- Heat capacity 3696 J/kg/K 12 reotactic swine brain atlas. The acoustic focus from a Thermal conductivity 0.55 W/m/K single-element, MR-compatible 1.14-MHz FUS transduc- Perfusion rate 14.1 kg/m3/sec er (1.5-mm FWHM laterally, 7.5-mm FWHM axially, and 5.7-cm focal distance, Fig. 1) was stereotactically aligned Baseline temperature 37 °C to the VPL (Fig. 2a and b). Acoustic energy was delivered Sonication sequence with the following parameters: 43.7-msec pulse duration, Initial sonication duration 0.0437 sec 10-Hz pulse repetition frequency, and 40-second total Number of additional pulse sequences 399 — sonication duration. The spatial acoustic intensity was 25 Duty factor 43.7 % 2 W/cm , and the mechanical index was 0.53 (Fig. 1c). Pulse cycle period 0.1 sec LIFU delivered to the right VPL decreased left median nerve SSEP amplitude to 71.6% ± 11.4% (mean ± SD) Cool-off duration 5 sec

878 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC Selective neuromodulation with LIFU

FIG. 2. Thalamic neuromodulation with LIFU. a and b: Axial and coronal T2-weighted MR images demonstrating the right VPL target (white), control left VPL target (black), and artifact from the cortical electrode (*). c: Representative median nerve SSEPs at baseline and at 5 minutes and 10 minutes following right VPL LIFU sonication and control, “off” target sonications. Note the sup- pression in SSEPs at 5 minutes after sonication, with a return to baseline by 10 minutes. d: SSEP amplitude changes following LIFU. Red arrow indicates time of LIFU. LIFU delivered to the right VPL decreased the left median nerve SSEP amplitude to 71.6% ± 11.4% (mean ± SD) compared with baseline recordings (4 animals, total of 9 sonications). SSEPs recovered to 87.2% ± 15.4% of baseline values within 10 minutes. Control sonications in the left VPL had no lasting effects on left median nerve SSEPs, which were 96.3% ± 7.4% of baseline values (n = 6 sonications, 4 animals). Error bars indicate SDs. Figure is available in color online only. to near baseline within 20 minutes (Fig. 2c and d). HIFU of baseline amplitude values (4 animals, 6 sonications; p ablations in the VPL resulted in permanent loss of SSEPs = 0.0002), and tibial-evoked potentials were not signifi- (Supplemental Fig. 1). cantly changed at 102.0% ± 4.3% compared with base- line values (4 animals, 6 sonications; p = 0.19) (Fig. 3f–h). Thalamic Mapping With LIFU Neuromodulation Then the acoustic focus was aligned with the right VPL Since the acoustic focus of the 1.14-MHz experimen- (Fig. 4i). LIFU sonication of the VPL had no significant tal ultrasound transducer and the 710- and 220-kHz clini- effect on trigeminal-evoked cortical potentials (103.9% ± cal ultrasound transducers (InSightec, Ltd.) were small 3.3%, p > 0.05), but tibial-evoked potentials decreased sig- nificantly to 83.9% ± 4.3% compared with baseline values enough to target a single thalamic nucleus in the swine -6 brain, we next sought to determine whether LIFU neuro- (4 animals, 6 sonications, p = 4.2 × 10 ) (Fig. 3j–l). modulation could be used to noninvasively map the ventro- lateral thalamic nuclear complex using the experimental (2 Temperature Monitoring During LIFU animals) and clinical (2 animals) ultrasound transducers. Next, we sought to determine whether the LIFU pa- Like the somatotopic organization of the , rameters used for thalamic neuromodulation caused tis- the ventrolateral thalamic nuclear complex has a distinct sue heating at the acoustic focus. We used 2D PRFS MR somatotopic organization with trigeminal (snout) inputs thermography to image the acoustic focus and measure synapsing in the ventroposteromedial (VPM) nucleus and tissue heating during LIFU and HIFU sonications.7 Dur- somatic (median and tibial nerve) inputs synapsing in the ing LIFU sonications (n = 4) there were no observable VPL nucleus. These nuclei are functionally distinct, but temperature increases at the acoustic focus in comparison are indistinguishable on high-resolution MRI, and histo- with background signal noise with a sensitivity < 5°C (Fig. logically they form a continuum without clear borders. 4b). Post LIFU T2*-weighted imaging did not show any To determine the spatial resolution of LIFU, we first signal change at the focus (Fig. 4a). During HIFU soni- sonicated a region of the thalamus 2 mm anterior to the cations (20 W, 20 seconds; 6 animals, 1 sonication per ventrolateral thalamic nuclear complex (Fig. 3a) with animal), peak voxel temperatures increased by amounts LIFU while recording evoked potentials from the tri- ranging from 25°C to 50°C (Fig. 4e), and post HIFU T2*- geminal (snout) and tibial (hindlimb) nerves in sequence weighted images demonstrated a small area of increased (Fig. 3b). During this “off-target,” control sonication, there signal at the acoustic focus corresponding to tissue abla- were no significant changes in the amplitude of either the tion (Fig. 4f). trigeminal-evoked or tibial-evoked cortical potentials To further determine whether LIFU sonications subtly compared with baseline recordings (trigeminal 98.2% ± increased tissue temperature by an amount that was be- 4.5% and tibial 100.6% ± 4.4%; p = 0.40 and 0.42, respec- low the sensitivity of standard MR thermography, we per- tively) (Fig. 3c and d, 4 animals, 5 sonications). Next, we formed a series of experiments using increasing sonication adjusted the acoustic focus to target the right VPM (Fig. duration and intensity. Using the same ultrasound param- 3e). LIFU sonication of the VPM resulted in a decrease eters described in Fig. 1c, we increased the total sonica- in trigeminal-evoked cortical potentials to 76.9% ± 7.5% tion duration 6-fold (240 seconds) to allow for increased

J Neurosurg Volume 128 • March 2018 879

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC R. F. Dallapiazza et al.

FIG. 3. Noninvasive thalamic nucleus mapping with LIFU. a, e, and i: Thalamic nucleus targeting for control (2 mm anterior to VPM, white shape), VPM (yellow shapes), and VPL (blue shapes) with LIFU. b, f, and j: Representative trigeminal and tibial SSEPs at baseline (black) and post-sonication (red). c and d: Nontarget control LIFU. The graphs show the mean right trigeminal and tibial (hindlimb) SSEP amplitude (c) as percentage of baseline value after nontarget, control LIFU and change in amplitude over time (d). The bars represent the composite mean SSEP amplitude for the 5 minutes after sonication. Error bars indicate SDs. Non- target, control LIFU did not alter either the trigeminal-evoked or tibial-evoked SSEP amplitude compared with baseline recordings (trigeminal 98.2% ± 4.5% and tibial 100.6% ± 4.4%; p = 0.40 and 0.42, respectively) (4 animals, total of 5 sonications). The acoustic focus was 2 mm anterior to the VPM target. g and h: VPM LIFU. Right VPM LIFU decreased the left trigeminal SSEPs to 76.9% ± 7.5% of baseline values (4 animals, total of 6 sonications; p = 0.0002), but the tibial SSEPs were not significantly changed at 102.0% ± 4.3% compared with baseline values (4 animals, total of 6 sonications; p = 0.19). k and l: VLP LIFU. Right VPL LIFU had no significant effect on trigeminal SSEP amplitude (103.9% ± 3.3%, p > 0.05), but tibial SSEPs decreased significantly to 83.9% ± 4.3% compared with baseline values (4 animals, total of 6 sonications, p = 4.2 × 10-6). Figure is available in color online only.

MR averaging and improved signal-to-noise ratios. During Histological Analysis - this sonication, average temperatures were 0.1°C ± 0.3°C Gross and microscopic histological analyses were per- compared with baseline temperatures (Fig. 4c). Next, we formed in all 10 animals. Eight animals received LIFU systematically increased the sonication energy from 0.25 targeting the right ventrolateral thalamus (Fig. 5a), and 6 W to 16 W. There were no observable temperature increas- animals received HIFU targeting the left thalamus (Fig. es with sonication intensities ranging from 0.25 to 2 W. 5e). Fixed whole brains were cut in the coronal plane, and Sonication at 4 W, a 16-fold increase in power, resulted in a thermal rise of ~ 3°C at the acoustic focus (Fig. 4d). the ventrolateral thalamic nucleus was identified grossly. We used the HIFU Simulator V1.2 to predict tempera- Among the animals that received LIFU in the right ven- ture changes based on our sonication parameters. Sonica- trolateral thalamus, there was no gross evidence of tissue tion parameters and beam path geometry were determined damage. In animals that received HIFU in the left thala- based on experimental settings, and tissue properties for mus, lesions were identified in all animals and varied in human gray matter were collected from published data appearance from dark discoloration without overt tissue (Tables 1 and 2). Resolution was adjusted to 1 mm to disruption to well-circumscribed lesions with surrounding match the resolution of MR thermography images. For tissue edema, which correlated to the peak temperature LIFU sonications, the predicted thermal rise at the focus observed during MR thermography. Serial sections were was 0.13°C (Fig. 4g), which agrees with our measured stained with H & E and LFB. In sections from animals thermal rise of 0.1°C ± 0.3°C. The predicted temperature treated with LIFU there was no microscopic evidence of rose steadily throughout the sonication, reaching a peak tissue damage. Neurons had a normal appearance, and at 40 seconds, and returned to baseline within 25 seconds there was no evidence of disruption of the (Fig. 4g). The predicted volume of tissue exposed to an tracts around the ventrolateral thalamic nuclei (Fig. 5b– increase in temperature of more than 0.05°C was 1 × 1 × d, LFB not shown). These were indistinguishable from 3 mm, which is in agreement with the dimensions of the control sections that were untreated with ultrasound. In acoustic focus (Fig. 4h). sections from animals treated with HIFU, there was a

880 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC Selective neuromodulation with LIFU

FIG. 4. LIFU temperature monitoring. a: T2-weighted image obtained after LIFU showing no change in signal at the acoustic focus. b: PRFS image showing no change in temperature during LIFU. c: Temperature changes at the acoustic focus measured every 5 seconds during a LIFU sonication with a 240-second duration showing a 0.1°C ± 0.3°C average temperature change. Blue points correspond to baseline MR images, and red points correspond to MR images acquired during sonication. s = seconds. d: Temperature changes with increasing sonication power confirming that there is no temperature increase until the acoustic power is increased by a factor of 16 (from 0.25 W to 4 W). e: T2*-weighted image obtained after HIFU showing small area of hyperin- tensity at the acoustic focus. f: PRFS image confirming temperature elevations during HIFU. g: Modeled temperature rise at the acoustic focus during LIFU sonication predicted by the HIFU Simulator V1.2. The predicted peak temperature rise, 0.13°C, occurs at sonication termination. h: Modeled thermal map showing predicted peak spatial temperature rise during LIFU with a 1 × 1 × 3 mm focal heating < 0.05°C. Values in color scales (b, d, and h) are °C. Figure is available in color online only. characteristic necrotic core with cellular debris and tis- sound focused on the motor cortex. However, a somato- sue disruption.10 Surrounding the necrotic core there were motor map has been difficult to assess due to a mismatch ischemic neurons and bubbly, edematous changes in the in the size of the acoustic focus and the size of the rodent white matter (Fig. 5f–h). Microscopic analysis of the over- brain.21,29,30,34 In human studies, Legon et al.24 found that ul- lying cortex and adjacent white matter tissues within the trasound targeting the postcentral could inhibit me- acoustic path of ultrasound did not show any evidence of dian-nerve evoked potentials, and translational movements tissue disruption. of the transducer 5 mm from the had no effect. However, this study did not examine whether lateral movements along the postcentral gyrus, corresponding to Discussion facial representation, could inhibit somatosensory evoked LIFU for Noninvasive Neuromodulation and Brain Mapping potentials from the trigeminal system. In these experiments, pulsed low-intensity focused ul- Current studies of low intensity ultrasound neuromodu- trasound (LIFU) was targeted to the ventrolateral thala- lation in large-brain animal models and humans have used mus in swine, resulting in reversible suppression of SSEPs short ultrasound pulses targeting the cerebral cortex. Def- by 20%–50%. Given the high spatial resolution of focused fieux et al.8 targeted the ( ultrasound, we were able to selectively target the subnuclei 8) in awake nonhuman primates and demonstrated im- within the somatosensory thalamus. Thus, we were able pairment in antisaccade task performance. Legon et al.24 to selectively suppress evoked potentials through the tri- also used short ultrasound pulses targeting somatosensory geminal system by targeting the VPM while leaving the cortex in concurrence with median nerve stimulation to lemniscal pathways that synapse in the VPL unaltered. demonstrate decreased evoked potential amplitudes re- Furthermore, focused ultrasound targeted as little as 2 corded by EEG. In each of these studies, ultrasound im- mm anterior to the ventrolateral thalamus had no effect on paired short-term neural function during brief tasks or evoked potentials. recording procedures with rapid recovery on subsequent Prior studies examining the effects of low-intensity trials. We found that longer ultrasound pulses delivered ultrasound in the brain have had limited success in dem- over prolonged periods of time resulted in a substantial onstrating anatomical specificity. Among rodent studies, and sustained decrease in neural function that lasted for a variety of motor responses have been elicited by ultra- a period of several minutes but caused no lasting damage.

J Neurosurg Volume 128 • March 2018 881

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC R. F. Dallapiazza et al.

FIG. 5. Gross and histological analysis of thalamic LIFU and HIFU. a–d: LIFU sonications demonstrating no evidence of histologi- cal damage. H & E (b–d); original magnification ×40 (b), ×100 (c), and ×400 (d). Black boxes represent the area depicted in the following panel. e–h: HIFU sonication showing well-circumscribed lesion in the ventrolateral thalamus. Microscopic analysis shows ischemic neurons along the periphery and edema extending into the white matter tracts. Boxes indicate the area depicted in the following panel. Figure is available in color online only.

LIFU Safety This concept not only applies to the plasma membrane The Food and Drug Administration regulates the safety but also to intracellular membrane compartments such as of diagnostic ultrasound devices, and several therapeutic synaptic vesicles and mitochondria, which rely on electro- ultrasound devices have investigational approval. Given the wide array of applications and device variability, there are several recommended safety metrics, including spatial average intensity, time-averaged intensity, and mechanical and thermal indices for diagnostic devices. These metrics can vary widely based on the acoustic frequency, acoustic power, duration of exposure, and pulsing schemes. Ulti- mately, the safety of LIFU devices will need to be deter- mined in animal models or in computational models. We monitored tissue temperature at the acoustic focus during LIFU sonications using MR thermography and found no heating above the sensitivity measurements (< 1°C). In addition, we applied the ultrasound parameters used for neuromodulation into HIFU simulation software and determined that theoretical temperature increases from these parameters would result in < 1°C heating. These results suggest that the ultrasound parameters used in this study are unlikely to significantly heat tissues and can safely be applied for human use.

Mechanism of LIFU Neuromodulation The precise mechanisms of neuromodulation with ultra- sound are incompletely understood. At higher intensities, focused ultrasound causes tissue heating and cavitation by generating frictional energy and low-pressure fields at the acoustic focus that can reversibly impair neural function. With LIFU, tissue heating and cavitation are less likely mechanisms. Focused ultrasound also exerts a mechani- 31 cal force as pressure waves move through tissues. One of FIG. 6. Thalamic LIFU neuromodulation occurs across ultrasound the leading hypotheses for low-intensity ultrasound neuro- frequencies. Representative trigeminal evoked potentials with LIFU modulation is that mechanical energy or intramembrane focused at the VPM nucleus demonstrating neuromodulation using 3 cavitation impart physiological changes in membrane flu- separate ultrasound transducers (220 kHz, 650 kHz, and 1.1 MHz). Fig- idity and permeability leading to neuronal depolarization. ure is available in color online only.

882 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC Selective neuromodulation with LIFU chemical gradients to properly function. There is already observed without evidence of tissue heating or histological evidence that low-intensity ultrasound can alter synaptic damage. LIFU is an emerging method for neuromodula- vesicle density shortly after exposure. tion and noninvasive brain mapping and may have a vari- Mechanical energy from ultrasound could also induce ety of applications in human brain studies and for thera- structural changes in transmembrane receptors and volt- peutic neuromodulation. age-gated ion channels that could lead to their activation or inhibition. Tufail et al.29 demonstrated that pharmaco- References logical inhibition of voltage-gated Na+ and Ca2+ channels 1. Adrianov OS, Vykhodtseva NI, Fokin VF, Uranova NA, Avi- blocks ultrasound-induced depolarization in both neurons rom VM: [Reversible functional shutdown of the optic tract and glia in ex vivo slice experiments. Furthermore, me- on exposure to focused ultrasound.] Biull Eksp Biol Med chanical energy from ultrasound could temporarily dis- 97:760–762, 1984 (Russian) rupt delicately organized synaptic signalosomes within 2. Chang WS, Jung HH, Kweon EJ, Zadicario E, Rachmilevitch the postsynaptic density. Adrianov et al.1 used electron mi- I, Chang JW: Unilateral magnetic resonance guided focused croscopy to analyze the effects of low-intensity ultrasound ultrasound thalamotomy for essential tremor: practices and clinicoradiological outcomes. J Neurol Neurosurg Psychia- on the feline lateral geniculate nucleus and found widen- try 86:257–264, 2015 ing of the synaptic cleft and a decrease in the thickness of 3. Clement GT, Hynynen K: A non-invasive method for focus- the postsynaptic density. Given the relatively long-lasting ing ultrasound through the human skull. Phys Med Biol effects we observed after LIFU sonication of the ventro- 47:1219–1236, 2002 lateral thalamus, it is tempting to speculate that prolonged 4. Clement GT, Sun J, Giesecke T, Hynynen K: A hemisphere ultrasound exposures could lead to some form of short- array for non-invasive ultrasound brain therapy and surgery. term synaptic plasticity. Certainly, further work is neces- Phys Med Biol 45:3707–3719, 2000 5. Clement GT, White J, Hynynen K: Investigation of a large- sary to support these claims; however, it bears mentioning area phased array for focused ultrasound surgery through the since long-term neural suppression with ultrasound could skull. Phys Med Biol 45:1071–1083, 2000 potentially have a large impact on noninvasive therapeutic 6. Clement GT, White PJ, King RL, McDannold N, Hynynen K: neuromodulation strategies. A magnetic resonance imaging-compatible, large-scale array for trans-skull ultrasound surgery and therapy. J Ultrasound Applications of LIFU Neuromodulation Med 24:1117–1125, 2005 7. De Poorter J, De Wagter C, De Deene Y, Thomsen C, Ståhl- In this study, we adapted an FDA-approved HIFU ultra- berg F, Achten E: Noninvasive MRI thermometry with the sound device and a second multi-element transducer (both proton resonance frequency (PRF) method: in vivo results in from Insightec) to apply LIFU to the thalamus in swine human muscle. Magn Reson Med 33:74–81, 1995 (Fig. 6). This required calibration of the device for lower 8. Deffieux T, Younan Y, Wattiez N, Tanter M, Pouget P, Aubry intensities and electronic programming to accommodate JF: Low-intensity focused ultrasound modulates monkey vi- pulsed (vs continuous) sonications. Thus, currently avail- suomotor behavior. Curr Biol 23:2430–2433, 2013 9. Elias WJ, Huss D, Voss T, Loomba J, Khaled M, Zadicario able HIFU systems can be easily modified to apply pulsed E, et al: A pilot study of focused ultrasound thalamotomy for LIFU. LIFU neuromodulation could be used immediately essential tremor. N Engl J Med 369:640–648, 2013 for intraprocedural target confirmation and brain mapping 10. Elias WJ, Khaled M, Hilliard JD, Aubry JF, Frysinger RC, during therapeutic, focused ultrasound ablations. For ex- Sheehan JP, et al: A magnetic resonance imaging, histologi- ample, LIFU could be targeted to the thalamic VIM nucle- cal, and dose modeling comparison of focused ultrasound, us to verify tremor suppression and to map the somatosen- radiofrequency, and Gamma Knife radiosurgery lesions in sory thalamus prior to ablation for patients being treated swine thalamus. J Neurosurg 119:307–317, 2013 11. Elias WJ, Lipsman N, Ondo WG, Ghanouni P, Kim YG, Lee for essential tremor. Subablative sonications are currently W, et al: A randomized trial of focused ultrasound thalamot- used during human HIFU thalamotomies for neuromodu- omy for essential tremor. N Engl J Med 375:730–739, 2016 lation to aid in determining the optimal ablation site; how- 12. Félix B, Léger ME, Albe-Fessard D, Marcilloux JC, Rampin ever, there are potential risks for permanent tissue damage, O, Laplace JP: Stereotaxic atlas of the pig brain. Brain Res even with moderate tissue heating. Since LIFU does not Bull 49:1–137, 1999 cause tissue heating, this method could be more useful for 13. Foley JL, Little JW, Vaezy S: Image-guided high-intensity extensive thalamic mapping during HIFU ablations. focused ultrasound for conduction block of peripheral nerves. Ann Biomed Eng 35:109–119, 2007 Furthermore, most current HIFU transducers are MR- 14. Fry FJ, Ades HW, Fry WJ: Production of reversible changes compatible, and focused ultrasound procedures are largely in the by ultrasound. Science 127:83– MR-guided. It is conceivable that LIFU neuromodulation 84, 1958 could be used in combination with MR-based measure- 15. Gavrilov LR: Use of focused ultrasound for stimulation of ments such as functional MRI. This could provide an nerve structures. Ultrasonics 22:132–138, 1984 unparalleled opportunity to study the human brain in a 16. Harvey EN: The effect of high frequency sound waves highly localized, noninvasive manner in a variety of neu- on heart muscle and other irritable tissues. Am J Physiol 91:284–290, 1929 rological diseases and among healthy individuals. 17. Heimburger RF: Ultrasound augmentation of central nervous system tumor therapy. Indiana Med 78:469–476, 1985 Conclusions 18. Hynynen K, McDannold N, Clement G, Jolesz FA, Zadicario E, Killiany R, et al: Pre-clinical testing of a phased array In this study, we demonstrate that LIFU can be used to ultrasound system for MRI-guided noninvasive surgery of the selectively inhibit thalamic relay neurons in swine with a brain—a primate study. Eur J Radiol 59:149–156, 2006 high spatial resolution. These physiological effects were 19. Ishihara Y, Calderon A, Watanabe H, Okamoto K, Suzuki Y,

J Neurosurg Volume 128 • March 2018 883

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC R. F. Dallapiazza et al.

Kuroda K, et al: A precise and fast temperature mapping us- continuum mechanics hypothesis. Neuroscientist 17:25–36, ing water proton chemical shift. Magn Reson Med 34:814– 2011 823, 1995 32. Tyler WJ, Tufail Y, Finsterwald M, Tauchmann ML, Olson 20. Jeanmonod D, Werner B, Morel A, Michels L, Zadicario E, EJ, Majestic C: Remote excitation of neuronal circuits using Schiff G, et al: Transcranial magnetic resonance imaging- low-intensity, low-frequency ultrasound. PLoS One 3:e3511, guided focused ultrasound: noninvasive central lateral thala- 2008 motomy for chronic neuropathic pain. Neurosurg Focus 33. Yoo SS, Bystritsky A, Lee JH, Zhang Y, Fischer K, Min BK, 32(1):E1, 2012 et al: Focused ultrasound modulates region-specific brain 21. King RL, Brown JR, Pauly KB: Localization of ultrasound- activity. Neuroimage 56:1267–1275, 2011 induced in vivo neurostimulation in the mouse model. Ultra- 34. Younan Y, Deffieux T, Larrat B, Fink M, Tanter M, Aubry sound Med Biol 40:1512–1522, 2014 JF: Influence of the pressure field distribution in transcranial 22. Lee W, Kim H, Jung Y, Song IU, Chung YA, Yoo SS: Image- ultrasonic neurostimulation. Med Phys 40:082902, 2013 guided transcranial focused ultrasound stimulates human primary somatosensory cortex. Sci Rep 5:8743, 2015 23. Lee W, Lee SD, Park MY, Foley L, Purcell-Estabrook E, Kim Disclosures H, et al: Image-guided focused ultrasound-mediated regional brain stimulation in sheep. Ultrasound Med Biol 42:459– Dr. Elias reports receiving support for non–study-related work 470, 2016 from InSightec and the Focused Ultrasound Foundation. 24. Legon W, Sato TF, Opitz A, Mueller J, Barbour A, Williams A, et al: Transcranial focused ultrasound modulates the activ- Author Contributions ity of primary somatosensory cortex in humans. Nat Neuro- Conception and design: Dallapiazza, Timbie, Holmberg, Price, sci 17:322–329, 2014 Miller, Elias. Acquisition of data: Dallapiazza, Timbie, Holmberg, 25. Lipsman N, Schwartz ML, Huang Y, Lee L, Sankar T, Chap- Gatesman, Lopes, Miller. Analysis and interpretation of data: man M, et al: MR-guided focused ultrasound thalamotomy Dallapiazza, Timbie, Holmberg, Lopes, Miller, Elias. Drafting for essential tremor: a proof-of-concept study. Lancet Neurol the article: Dallapiazza, Timbie, Gatesman, Lopes, Miller, Elias. 12:462–468, 2013 Critically revising the article: all authors. Reviewed submitted 26. Magara A, Bühler R, Moser D, Kowalski M, Pourtehrani version of manuscript: all authors. Approved the final version of P, Jeanmonod D: First experience with MR-guided focused the manuscript on behalf of all authors: Dallapiazza. Statistical ultrasound in the treatment of Parkinson’s disease. J Ther analysis: Dallapiazza, Timbie, Miller. Administrative/technical/ Ultrasound 2:11, 2014 material support: Gatesman. Study supervision: Dallapiazza, 27. Meyers R, Fry WJ, Fry FJ, Dreyer LL, Schultz DF, Noyes RF: Price, Elias. Early experiences with ultrasonic irradiation of the pallidofu- gal and nigral complexes in hyperkinetic and hypertonic dis- Supplemental Information orders. J Neurosurg 16:32–54, 1959 Online-Only Content 28. Pinton G, Aubry JF, Bossy E, Muller M, Pernot M, Tanter M: Attenuation, scattering, and absorption of ultrasound in the Supplemental material is available with the online version of the skull bone. Med Phys 39:299–307, 2012 article. 29. Tufail Y, Matyushov A, Baldwin N, Tauchmann ML, Supplemental Fig. 1. https://thejns.org/doi/suppl/10.3171/2016.​ Georges J, Yoshihiro A, et al: Transcranial pulsed ultrasound 11.JNS16976. stimulates intact brain circuits. Neuron 66:681–694, 2010 30. Tufail Y, Yoshihiro A, Pati S, Li MM, Tyler WJ: Ultrasonic Correspondence neuromodulation by brain stimulation with transcranial ultra- Robert F. Dallapiazza, Department of Neurosurgery, University sound. Nat Protoc 6:1453–1470, 2011 of Virginia, School of Medicine, Box 800212, Charlottesville, VA 31. Tyler WJ: Noninvasive neuromodulation with ultrasound? A 22908. email: [email protected].

884 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 10/04/21 01:36 PM UTC