If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Hvidovre, DenmarkCenter for Magnetic Resonance, Department of Health Technology, Technical University of Denmark, Kgs, Lyngby, Denmark
Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Hvidovre, DenmarkCenter for Magnetic Resonance, Department of Health Technology, Technical University of Denmark, Kgs, Lyngby, Denmark
Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Hvidovre, DenmarkDepartment of Neurology, Copenhagen University Hospital Bispebjerg, Copenhagen, DenmarkInstitute of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
Corresponding author. Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Hvidovre, Section 714, Kettegaard Allé 30, 2650, Hvidovre, Denmark.
Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Hvidovre, DenmarkCenter for Magnetic Resonance, Department of Health Technology, Technical University of Denmark, Kgs, Lyngby, Denmark
TFUS is an emerging non-invasive stimulation method with excellent focality
•
We summarize the safety-related data from the available literature.
•
Adverse effects were absent in 30 studies, and were reported in 3 studies.
•
Many studies used parameters outside safety limits for diagnostic ultrasound.
•
Further studies are warranted to establish the safety margin for TFUS.
Abstract
Background
Low-intensity transcranial focused ultrasound stimulation (TFUS) holds great promise as a highly focal technique for transcranial stimulation even for deep brain areas. Yet, knowledge about the safety of this novel technique is still limited.
Objective
To systematically review safety related aspects of TFUS. The review covers the mechanisms-of-action by which TFUS may cause adverse effects and the available data on the possible occurrence of such effects in animal and human studies.
Methods
Initial screening used key term searches in PubMed and bioRxiv, and a review of the literature lists of relevant papers. We included only studies where safety assessment was performed, and this results in 33 studies, both in humans and animals.
Results
Adverse effects of TFUS were very rare. At high stimulation intensity and/or rate, TFUS may cause haemorrhage, cell death or damage, and unintentional blood-brain barrier (BBB) opening. TFUS may also unintentionally affect long-term neural activity and behaviour. A variety of methods was used mainly in rodents to evaluate these adverse effects, including tissue staining, magnetic resonance imaging, temperature measurements and monitoring of neural activity and behaviour. In 30 studies, adverse effects were absent, even though at least one Food and Drug Administration (FDA) safety index was frequently exceeded. Two studies reported microhaemorrhages after long or relatively intense stimulation above safety limits. Another study reported BBB opening and neuronal damage in a control condition, which intentionally and substantially exceeded the safety limits.
Conclusion
Most studies point towards a favourable safety profile of TFUS. Further investigations are warranted to establish a solid safety framework for the therapeutic window of TFUS to reliably avoid adverse effects while ensuring neural effectiveness. The comparability across studies should be improved by a more standardized reporting of TFUS parameters.
Weak Transcranial Focused Ultrasound Stimulation (TFUS) aims to modulate neural activity by delivering a focused ultrasonic beam to a small target area in the brain. Currently, interest in TFUS is strongly increasing as it holds the promise of a far better spatial resolution than established non-invasive stimulation techniques and of the ability to reach deep brain areas [
]. This might open up intriguing new applications such as epilepsy treatment or pre-surgical diagnostics prior to electrode implantation for deep-brain stimulation [
]. TFUS is also attractive because it can be readily combined with neuroimaging modalities such as functional magnetic resonance imaging (fMRI) and electroencephalography (EEG) without interfering with the recordings, as it applies acoustic waves rather than electric or magnetic fields.
Firmly establishing its safety profile is a central requirement when aiming to move TFUS from initial pilot studies towards broader testing in humans in-vivo. Reviews on safety and bio-effects of ultrasound (US) in diagnostics [
] are available and constitute a benchmark to avoid harmful effects also for TFUS. Relating the TFUS parameters to these guidelines, as done in many of the published studies, might be considered a conservative choice. However, several aspects put TFUS in a special position. TFUS usually employs lower frequency compared to diagnostic ultrasound (usually upper kHz range vs. MHz) and longer pulse bursts. TFUS has a static focus so that the total energy delivered at the focal point can be higher than the maximal local energy deposit for diagnostic US, as the latter uses scanning approaches. The mechanism-of-action of TFUS is still poorly understood, rendering it more difficult to principally exclude harmful effects. In addition, current findings about the dose-response curve of TFUS [
] suggest that future therapeutic applications might aim to use intensities above the safety limits for diagnostic US in order to increase the robustness of the neural effects. Such a choice requires solid knowledge about the safety margin of TFUS. Along similar lines, accurate dose control for human TFUS is complicated by the presence of the skull, which strongly attenuates the beam. The attenuation depends on the individual skull thickness and composition [
], which are difficult to account for and lead to conservative intensity choices with an increased risk of underdosing. If the safety margin of TFUS is not well established, the use of more lenient dosing strategies to mitigate this problem is not feasible.
There is a pressing need to establish specific safety guidelines for TFUS. Yet, the current knowledge about the risk-benefit ratio and the therapeutic window of TFUS is still rudimentary because TFUS is at an early stage of development. Indeed, no dedicated phase I safety human study has been performed so far, but the safety profile needs to be systematically investigated and monitored to ensure the patients’ safety. However, relevant information is already available today, because some of the published studies on TFUS in animals or humans included safety-relevant tests. Here, we systematically summarize these findings to give an overview of the current state of knowledge about TFUS safety. We start by describing the relevant physical parameters used to characterize the TFUS stimulus. We then shortly describe the known physical mechanisms by which ultrasound can cause tissue damage and we introduce the established safety indices, based on the beam parameters. Finally, we introduce the methods that have so far been applied to test for adverse effects of TFUS, and list the corresponding results. In the discussion, we summarize the implications of the available findings for in-vivo human TFUS applications.
Material and methods
Literature review on the safety of TFUS
For this systematic review, we followed the PRISMA guidelines [
The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration.
]. Details on the implementation of the PRISMA requirements in our review are stated in the Supplementary material (Table S1). Our review was based on searches in PubMed (www.ncbi.nlm.nih.gov/pubmed) and bioRxiv (https://www.biorxiv.org/) for published and pre-published studies, using the keywords ‘tFUS’, ‘LIFUP’, ‘noninvasive brain stimulation focused ultrasound’, ‘neuromodulation brain transcranial ultrasound’, ‘focused ultrasound transcranial brain stimulation’ and ‘pulsed ultrasound brain stimulation’. The eligibility criteria were low intensity, low frequency TFUS in the brain of animals or humans with safety assessment, without use of microbubbles. Additional sources were reviews of the literature lists of relevant papers, and papers pointed out by the reviewers during the peer-review process. Fig. 1 shows details of the literature search. The last complete search was performed in January 2019 by one of the authors, and the last update was done in June 2019. From each paper, the sonication parameters and the methods used to assess safety and adverse effects were extracted as shown in Table 2 and Table 3 and categorized as described further below. Often, only some of the safety indices were reported. In that case, we give estimated values when possible.
Fig. 1Selection process for the studies included in this review. The scheme is from Refs. [
The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration.
Table 1Allowed limits for MI, TI, Ispta and Isppa according to the FDA guidelines for diagnostic ultrasound. The limit for TI also applies to TIC when bone is close by.
Table 2Overview of the parameters used in the reviewed studies. Ispta values very often exceeded the limits for diagnostic US. Cases where the Ispta values were higher than 3 W/cm2, corresponding to the limit for physiotherapeutic US, are highlighted in bold. Also one case in which MI exceeded the limit of 1.9 is marked in bold. If needed, we calculated missing parameters from the available data stated in the paper, which we indicate by “*” in the table. When the peak pressure was reported, MI was calculated using its definition (eq. (2)), and Isppa in water as indicated in Fig. 1 (ρ = 1000 kg/m3, c = 1500 m/s). Ispta was finally determined as Isppa x DC. 1) For [
]. 5) Not clear if it is in water or after cranial transmission. 6) A spatial average of intensity of 25–30 W/cm2 is used. 6) Modulated focused ultrasound means that two transducer, one driven at 2.25 MHz and the other at 1.75 MHz, producing a difference frequency at 500 kHz at the focus, and a carrier frequency of 2 MHz.
Study
Target
Parameters
Observed neural effect and adverse effect (if any)
Neuron's spike frequency and c-fos+ cell density increase and the activity of endogenous brain-derived neurotrophic factor (BDNF) were stimulated. Low frequency (250 KHz) and low intensities (up to around Ispta = 80 mW/cm2) result in more robust EMG response. The EMG failure probability increased with shorter ISI (200 ms), but decrease with multiple stimuli. BBB intentionally opened with the use of microbubbles.
fc = 650 kHz and Isppa in the range 0.5–20 W/cm2 did not elicit eye movement in any animals. Movements observed when fc = 350 KHz for an Isppa of 8.6 W/cm2.
100 (groups of sonications repeated up to 8 times per animal)
5 s (motor cortex) or 1 s (visual cortex)
in the range 0.5–1.4 after cranial tx
Up to 11.8 W/cm2 after cranial tx –SM1 Up to 14.3 W/cm2 after cranial tx -V1
Up to 5.9 W/cm2 * after cranial tx –SM1 Up to 7.15 W/cm2 * after cranial tx -V1
MEP or VEPs were detected over a certain intensity threshold, which varied across sheep and was always above diagnostic limits, and in some cases also above the physiotherapy limit. In both cases, higher Isppa result in stronger response amplitude. Four animals which underwent 600 sonications at Isppa = 6.6–10.5 W/cm2 showed micro-hemorrhages in the primary visual cortex.
Rabbit (after craniotomy), SM and visual area (the bottom line is only for temperature increase study)
690
0.05, 0.5, 10 and 50 ms
10, 20, 100 and 1000 Hz
0.5, 1, 1.5, 2, 9 s
1
–
<0.5 in water (for an Isppa = 3.3 W/cm2, resulting in clear BOLD activity)
3.3, 6.4, 9,5, 12.6 W/cm2 in water
1.6 W/cm2 in water (for Isppa = 3.3 W/cm2)
The BOLD activation was observed at a much lower acoustic intensity (Isppa = 3.3 W/cm2, Ispta = 1.6 W/cm2) compared to the intensity that resulted in forepaw movement (Isppa = 12.6 W/cm2, Ispta = 6.3 W/cm2)
690
0.5 ms
100 Hz
27 s
1
–
?
23 W/cm2
1.15 W/cm2
Parameter tested as control for temperature increase
0.62 after cranial tx (maximal simulated value across N = 12 subjects)
3 W/cm2 in water 2.5 W/cm2 after cranial tx (maximal simulated value)
1.5 W/cm2 in water 1.25 W/cm2 * after cranial tx (maximal simulated value)
Tactile sensations were not the same among subjects, but mostly at the hand area contralateral to the sonicated hemisphere. 1 out of 12 subjects did not report any sensation. Different peak amplitudes of EEG recording of SEP with and without stimulation.
35 W/cm2 in water <8.8 W/cm2 after cranial tx (estimated)
17.5 W/cm2 in water < 4.4 W/cm2 after cranial tx (estimated)
Response rates of elicited sensations during the FUS procedures were different among subjects (68 ± 28% S1, 59 ± 22% S2, 61 ± 26% S1+S2, average ± sd across subjects).
22.4 W/cm2 after cranial tx (max value reported, corresponding to animal with signs of bleeding)
11.2 W/cm2 after cranial tx (max value reported)
Motor responses were observed at minimum threshold (Isppa = 4.9–5.6 W/cm2, Ispta = 2.5–2.8 W/cm2) in a limited range of sonication parameters (TBS = 1–5 ms, 50% of duty cycle, and SD = 300 ms, at fc = 350 kHz). Pulsed sonication elicited motor responses at lower acoustic intensities than its equivalent continuous sonication (Isppa = 7.73 W/cm2). One animal which underwent a sonication of Ispta = 11.2 W/cm2 for a short period of time (<9 s using 1 ms TBD, 50% duty cycle and 300 ms SD) showed signs of local bleeding.
2.8 * in water 1.2 after cranial tx (maximal simulated value across N = 19 subjects)
16.6 W/cm2 in water 11.6 W/cm2 after cranial tx (maximal simulated value)
8.3 W/cm2 in water * 5.8 W/cm2 * after cranial tx (maximal simulated value)
fMRI: 11 out of 19 participants reported the perception of phosphenes, and a clear fMRI response. EEG: 10/10 subjects reported phosphene sensation. Changes in VEP EEG peak.
The sonication reduced the time to emergence of voluntary movement from intraperitoneal ketamine-xylazine anesthesia. A preliminary test showed that a Isppa = 3.3 W/cm2 failed to decrease the duration of the anesthetic state.
Amplitudes of SEPs (recorded by EEG) elicited by median nerve stimulation were significantly attenuated. The spectral content of sensory-evoked brain oscillations were significantly modulated by tFUS.
The amplitude of single-pulse TMS MEPs was decreased; the intracortical facilitation was attenuated; no effect on intracortical inhibition. Ultrasound reduces reaction time on a simple stimulus response task
Minimum value: 2.1 W/cm2; incremented by 1 W/cm2; maximum value: 14.9 W/cm2 5)
7.5 W/cm2 5)
Different thresholds to evoke observed motor response: Isppa = 3.4 ± 1.8 W/cm2 for the awake condition (grand mean response rate 76.2%) Isppa = 10.2 ± 2.4 W/cm2 (grand mean response rate 68.6%) or 12.4 ± 2.8 W/cm2 (grand mean response rate 38.6%) for 2 different types of anesthetics5)
1.87 in water 1.08 after cranial tx (estimated after measurement on skull attenuation)
29.5 W/cm2 in water 9.9 W/cm2 after cranial tx
1.34 W/cm2 in water 0.452 W/cm2 after cranial tx
Excitation effects with BOLD fMRI not only at the target but also off-target somatosensory and associated brain regions as a cause of modulation in downstream brain regions.
AEP decreases by 59.8 ± 3.3% (with Isppa = 2.3 W/cm2) and by 36.9 ± 7.5% (with Isppa = 4.6 W/cm2) of the baseline value in rats. AEP amplitudes decreased to an average of 27.7 ± 5.9% of baseline in pigs. This effect lasted between 30 min and 1 month in most treated animals.
Isppa = 1 W/cm2, TBD = 0.5 at PRF = 100 Hz and Isppa = 3W/cm2, TBD = 0.5 ms, PRF = 20 Hz, corresponding to 50 and 30 mW/cm2 Ispta did not change VEP. Isppa = 3 W/cm2 with TBD = 0.5 ms and PRF = 100 Hz (5% duty cycle) successfully suppressed the VEP. Higher duty cycle (8.3%) increased the VEP. The same effect was observed at Isppa = 5 W/cm2 and 5% duty cycle.
Enhancement of sensorimotor recovery after stroke. Decreased level of brain edema and tissue swelling in the affected hemisphere 3 days after the stroke.
Monkey amygdala and anterior cingulate cortex (ACC)
250
30 ms
10 Hz
40 s
1
–
Maximum 2.64 in amygdala and 1.64 in ACC * (from estimation after cranial transmission)
Maximum 51 W/cm2 in amygdala and 17 W/cm2 in ACC (estimation after cranial transmission)
Maximum 15.3 W/cm2 in amygdala and 5.3 W/cm2 in ACC (estimation after cranial transmission)
After TFUS, the functional coupling of the stimulated areas, but not of control areas, was selectively reduced. This effect was measured by fMRI and lasted for more than 1 h after stimulation.
Extracellular GABA level started to decrease upon sonication and remained reduced compared to control group up to 100 min after the end of sonication. The same effect was not observed for the extracellular glutamate level.
The robustness of the visual observed responses increased and the latency of the response decreased with increasing Ispta. Ispta = 3.46 w/cm2 was sufficient to induce strong motor response; no response was observed for Ispta<1.16 W/cm2. Ultrasound-induced motor responses were inhibited more than 20 min after ketamine injection. This was confirmed in in vitro cortical neuron sample by fluorescence calcium imaging, showing a dose-dependent effect.
US directly evokes hindlimb movement, even at short burst (3 ms) and had short latency (10 ms) and long refractory (3 s) periods. US modulation significantly suppressed forelimb and hindlimb responses following ECS for several minutes after the stimulation, but shows no short-term effect.
Rat cortex (target to elicit motor response, not corresponding to motor cortex)
320
0.23 ms
2 kHz
250 ms
?
10 s
From 0.7 to 1.77 *
Isppa of 7.5 W/cm2 (to have 50% response) in water. Via computer stimulation, it corresponds to 17.5 W/cm2 after cranial transmission due to reverberation
3.75 W/cm2 in water and 8.75 W/cm2 after cranial transmission * (to have 50% response)
A pressure threshold of 0.79 and 0.59 MPa was required to reach 50% of responsiveness, for deep or light anesthesia stage, respectively, and the sigmoid respond was less sharp in the light anesthesia stage. These pressures corresponded to an average Isppa of 7.5 W/cm2.
500 (from unfocused ultrasound or modulated focused ultrasound 6), mFUS)
0.2 ms
1.5 kHz
10 s
1
–
?
0.45–16 W/cm2 for unfocused US * 3–33 W/cm2 for mFUS *
0.15–5.25 W/cm2 for unfocused US 1–10 W/cm2 for mFUS
Increasing the Ispta increase the motor movement robustness, assessed by visual assessment with unfocused US and mFUS, and the normalized success rate in mFUS.
Table 3Overview of the safety assessments included in the reviewed studies. The involved methods are: fluorescein isothiocyanate-dextran (FITC–Dextran), trypan blue dye (T.b.), Evans blue dye (E.B.), magnetic resonance contrast agent (MR c.a.) to assess the BBB opening; antibodies to Caspase-3, quantitative transmission electron microscopy (e.m.), hematoxylin and eosin (H&E), terminal deoxynucleotidyl transferase dUTP Nick-End Labeling (TUNEL) assay, cresyl violet (c.v.), GFAP (glial fibrillary acidic protein), VAF (Vanadium acid fuchsin) and luxol fast blue dye (LFB) to monitor cell death, damage, brain ultrastructure and hemorrhage; Sensors (thermo-couple [
]), maximum temperature increase, (equation (7) to estimate ), magnetic resonance thermometry (MR th.) and the bioheat equation for the temperature increase; motor task (m.t.) or other for behavioral assessments. 1) safety assessed in rats that did not undergo pentylenetetrazol (PTZ) injection to induce epileptic activity [
]. The initial hypotheses for the ultrasound neuromodulation were thermal effects and acoustic cavitation. While an increase in the tissue temperature could perturb neuronal activity levels, the temperature increase due to low-intensity ultrasound is often less than 0.1 °C. Thus, the thermal effects of low-intensity ultrasound are most likely negligible. The second hypothesis is based on acoustic cavitation. This hypothesis postulates that the ultrasound generates nanobubbles in the lipophilic zone of the plasma membrane, which then vibrates according to the pressure variations, alters the local curvature of the bilayer, and changes overall neuronal excitability [
]. However, since nanobubbles are formed at an intensity larger than 100 mW/cm2, generation of micro or nanobubbles at the intensity used in standard neuromodulation protocols must be confirmed. The recent hypotheses now focus more on the effects of acoustic radiation forces on the permeability of the ion channels, such as mechanosensitive channels [
]. Another kind of hypotheses includes plasma deformation, which postulates that vibration of surrounding extra- and intracellular environment evokes mechanical changes in either the plasma membrane tension or the lipid bilayer and modulates neuronal activities [
Contrary to these works on the mechanisms involved with direct modulation of ion channels and membranes, an indirect in vivo ultrasound neuromodulation through auditory or cochlear pathways has been also recently proposed [
]. These studies demonstrated that ultrasound-induced activities were eliminated or reduced upon transection of the auditory nerves or removal of cochlear fluids. These results raised an important question of whether direct activation of neurons in the intact brain is possible. While more in-depth studies on the experimental protocols such as sharpness of the pulse, pulse repetition frequency, and bone transduction must be performed, these studies underscore the need for a solid understanding of the underlying mechanism of ultrasound neuromodulation [
Physical parameters and safety indices of US waves
A sketch of an experimental setup for TFUS is shown in Fig. 2A, using the stimulation of a rat as example. The main indices used to assess safety are:
•
(spatial peak temporal average intensity) is the temporal average intensity, calculated at the position of the spatial maximum
•
(spatial peak pulse average intensity) is the pulse average intensity, calculated at the position of the spatial maximum
•
(mechanical index) gives an estimation of the likelihood of inertial cavitation
•
(thermal index) is the steady-state temperature increase in soft tissue during ultrasound sonication
•
(thermal index for cranial bone) is a modification of TI, when the skull is close to the transducer face
Fig. 2Overview of the TFUS setup and parameters. A) The ultrasound pressure wave is generated by a transducer and delivered to the target through a guide filled with acoustic gel. B) The pressure stimulus over time is shown to indicate the main parameters. C) The main intensity values are shown for a fixed space position, together with their relationship with the pressure signal.
Ispta, TI and TIC are related to the risk of thermal bio-effects, while Isppa and MI are related to the risk of cavitation. The upper limits for these five indices allowed for diagnostic ultrasound are shown in Table 1. It should be noted that another guideline, IEC standard 60601-2-5 for physiotherapy US equipment, sets an upper limit for the “effective intensity’’, defined as the ratio of acoustic output power to effective radiating area, of 3 W/cm2. The standard also states that this value should only be reached for short times to prevent substantial heating. The “effective intensity’’ of 3 W/cm2 is usually interpreted as the upper limit for Ispta [
] compare the intensities used in their study against this limit rather than using the FDA guidelines for diagnostic US. Complementary to TI, the temperature increase at the target can be calculated as (equation 7 and 8 in Supplementary material) or through the bio-heat equation [
]. A more detailed explanation of these indices and formulae can be found in the Supplementary Material.
Mechanisms underlying tissue damage by US
Ultrasound waves may cause harmful effects on tissues via two physical mechanisms, mechanical and thermal. The main mechanical effect is cavitation, in which vapor cavities (or “bubbles”) form in the soft tissues during the periods of low pressure (i.e. the minima) of the acoustic wave cycles. Depending on intensity and center frequency, this can result in a stable oscillation (stable or non-inertial cavitation) or can result in violent bubble collapses (inertial cavitation) that create large forces in their neighborhood. The air bubbles can have an endogenous origin (for example in the lungs or intestine), or they can be created by the mechanical wave itself, if the peak rarefaction pressure (i.e. the pressure during the minima) is small enough to allow the liquid to reach vaporization. Alternatively, ultrasound contrast agents (UCA), which contain microbubbles, can be injected for, e.g. clinical purposes [
When a mechanical wave propagates linearly in a medium, its amplitude decreases exponentially starting from the source. The attenuation is caused by both scattering, i.e. the change in the direction of wave propagation due to the presence of microscopic obstacles along the beam, and absorption. Absorption is the process by which the wave energy is converted into heat, and therefore the medium is heated. Several ways to model or monitor the resulting temperature increase in the medium exist, and they will be further discussed below.
Types of adverse and side effects caused by TFUS
In this section, we summarize the potential adverse and side effects, which have so far been tested in TFUS studies, and briefly outline the employed techniques to assess the occurrence of these effects. The majority of results were obtained in animal studies, which tested for the following effects:
•
Blood-brain barrier (BBB) opening: The BBB is a semi-permeable membrane formed by endothelial cells which separates the vessels and the central nervous system (CNS) [
]. Air bubbles subjected to cavitation can break the BBB. Exploiting this effect, TFUS combined with US contrast agents is tested as a method for targeted drug delivery [
]. However, BBB opening is undesired for normal TFUS. Assessing BBB integrity is usually based on the intravenous injection of a substance, which cannot cross the barrier under normal conditions, prior to sonication. It is then tested whether TFUS causes the substance to diffuse into brain tissue. The dyes fluorescein isothiocyanate-dextran (FITC-dextran) [
] have been used for this purpose, and their presence inside the brain was investigated in post-mortem microscopy analyses of brain slices. Alternatively, an MRI contrast agent (a gadolinium chelate) was injected before the stimulation and its penetration into brain tissue was tested by assessing the MRI signal change due to the contrast agent [
], used to identify myelin in nervous tissue. Cell death can be of two types, apoptosis and necrosis. While apoptosis is part of the normal life cycle of the cells, necrosis is harmful and triggered by external factors or disease. It is possible to differentiate between both types of cell death based on morphological criteria, but this requires experience [
]. Additional techniques specifically label apoptotic cells and have therefore been used to distinguish between apoptosis and necrosis. For example, the presence of fragmented DNA is a sign of apoptosis, and not necrosis, and it can be labeled by terminal deoxynucleotidyl transferase (TdT) dUTP Nick-End Labeling (TUNEL) assay [
]. Acidophilia refers to the property of cells of staining readily with an acid dye and occurs after acute neuronal damage and death in brain ischemia.
Finally, also transmission electron microscopy has been used to quantitatively observe the effect of ultrasound on brain ultrastructure (postsynaptic density, docked vesicles, etc.) [
Irreversible changes of neural activity: Recordings after the sonication can determine whether changes in neural activity are reversible and characterize the duration of recovery. The effects on local neural activity in the TFUS target region can be detected directly via invasive recordings or voltage sensitive dyes [
]. TFUS-related changes in extracellular concentrations of excitatory and inhibitory neurotransmitters such as glutamate and λ-aminobutyric acid (GABA) can be measured via microdialysis techniques [
]. TFUS has also been combined with measurements of the forelimb and hindlimb responses to epidural cortical stimulation (ECS) to assess the cortical excitability changes after sonication [
] give non-invasive but less specific measurements of neural activity changes.
•
Undesired changes in animal behavior: TFUS may affect normal behavior in unintended ways. In animals, this is controlled by monitoring of every day behavior, like food uptake, defecation and movement behavior and checking for signals of pain and distress or change in weight [
]. In addition, tasks such as the rotorod task and wire-hanging task allow for quantitatively assessing the impact of TFUS on specific aspects of behavior [
]. One study induced ischemic stroke in mice and compared the behavioral changes of the mice which were treated with TFUS via a balance test and an adhesive removal test [
Adverse effects can be caused by cavitation or tissue heating. As outlined above, cavitation is prevented by controlling the pressure levels. While the temperature increase in the brain can be roughly estimated using Equations (7) and (8) in the Supplemetary Material [
] in the brain of the animal after craniotomy to track the temperature change in real time during sonication. A non-invasive alternative to this approach is measuring the temperature increase with thermocouples in a phantom [
So far, most TFUS studies used animal models. Tests for adverse effects in the few human studies were based on neurological examinations and/or structural MR imaging before and at one or several time points after the experiment [
]. In some of the studies, the participants were additionally contacted by telephone 2 months after the experiment and interviewed about any changes in their mental and physical health status, including experiences of any discomfort [
] presents results of phone interviews based on a ‘Participant report of symptoms questionnaire’ of 64 participants who had participated in one or more of seven human TFUS experiments before.
Results
Studies screened in this review
This systematic review follows PRISMA guidelines [
The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration.
], and the PRISMA checklist can be found in the Supplementary Methods (Table S1). The reviewing process shown in the PRISMA diagram Flow (Fig. 1) resulted in the selection of 31 peer-reviewed and 2 pre-published studies included in this review. From each of those papers (a complete list with citations is shown in Table 2), the sonication parameters (Table 2) and the methods used to assess safety and adverse effects (Table 3) were extracted and categorized as described further below. Often, only some of the safety indices were reported. In that case, we give estimated values when possible. The risk of bias was assessed and is reported in a separate section in the Supplementary Material.
BBB opening
BBB opening did not occur in any of the included studies [
] tested different sonication parameters on eight sheep in total, and reported four animals with micro-hemorrhages in the primary visual cortex after undergoing 600 sonications at 6.6 W/cm2 Isppa (6 repetitions of 100 sonications, with 30 s gaps). While the reported value for Isppa is within FDA limits, our calculated value for Ispta of 3.3 W/cm2 is exceeding the diagnostic limit, and is also slightly higher than the limit for physiotherapeutic US of 3 W/cm2. Interestingly, a sheep undergoing a single sonication at an Isppa of 13.4 W/cm2 did not present micro-hemorrhages. In another study [
], 1 of 37 rats was exposed to a high intensity (11.2 W/cm2 Ispta) for a short period of time (<9 s using 1 ms TBD, 50% duty cycle and 300 ms SD). It exhibited several areas containing hemosiderin, which indicate the potential of local bleeding, while none of the other animals showed any sign of bleeding.
Cell damage or death
Most of the studies testing for cell damage or death [
] observed no differential GFAP expression between the control and sonicated hemisphere for Isppa = 0.69 W/cm2, suggesting the absence of neural trauma. However, an increased number of astrocytes was observed for a control condition with Isppa = 280 W/cm2 (∼1.5 times above the FDA limit). Interestingly, no damage was observed even when AEP was not fully recovered after one month [
] tested parameter ranges for excitatory and inhibitory TFUS effects in craniotomized rats. While excitatory effects were very short-termed, suppression effects lasted several minutes. A reduction in the EEG response of up to 80% and a corresponding reduction of the BOLD signal that both lasted up to 10 min were reported for a long sonication duration of 9 s. Dallapiazza et al. [
] observed peak electrophysiological suppression in SSEP 5 min post-treatment, and the values returned to near baseline within 20 min. A further study [
] tested the facilitatory effects of ultrasound on somatosensory evoked potentials by measuring the changes in fractional fluorescence in the brains of mice dyed with voltage sensitive dyes. The TFUS-related changes disappeared within 20 min after ultrasound stimulation. Yang et al. [
] observed a decreased extracellular GABA level (approximately 20% below baseline) compared to a control group that lasted up to 100 min after the sonication ended. The same effect was not observed for glutamate. Gulick et al. [
] showed that TFUS significantly suppressed forelimb and hindlimb responses to ECS for several minutes after the stimulation blocks, even though effects immediately after single, short TFUS trials were absent. Kim et al. [
] observed a local increase in glucose metabolism induced by FUS to rat brain. This effect was demonstrated via PET imaging, which was started 20 min after the sonication and performed for 1 h. After that time, the metabolism had still not returned to baseline. In a work [
] on primates, the authors observed change in functional connectivity after a long sonication of 40 s at Ispta = 7 W/cm2. The change lasted for more than 1 h after sonication. A similar effect was observed in a related work [
] observed that the SEP signals after 10 min sonication were distinctively different compared to the control condition, even 35 min after the sonication. Daniels et al. [
] observed a full recovery of AEP amplitudes in rats within maximum 1 week post-treatment with Isppa = 2.3 W/cm2, while the signal from 5 out of 10 rats recovered up to one month post-treatment for an Isppa = 4.6 W/cm2. In the same study, 1 out of 5 pigs showed a fully recovered signal 1 h post-treatment while the other did not show any recovery 3 h post-treatment (in all 5 cases, Isppa = 4.6 W/cm2). Kim et al. [
] observed an increase in VEP in rats up to 5 min post-treatment with Isppa = 5 W/cm2 and a slight increase of VEP 150 s after treatment when Isppa = 3W/cm2. One study [
] induced ischemic stroke in mice and found a better sensorimotor performance in mice that underwent 20 min TFUS session via a balance test and an adhesive removal test. These improvements lasted for 4 weeks after treatment, suggesting an enhancement in brain plasticity. In the same study, the TFUS treatment in cerebellar LCN significantly lowered the percentage change in increased water content and tissue swelling in the ipsilateral hemisphere to the stroke.
Animal behavior
Several studies tested for changes from normal daily behavior after the sonication studies [
] (rotorod running task and wire-hanging task), without revealing differences in motor performance.
Temperature
Theoretical calculations based on Equations (7) and (8) in the Supplementary Methods suggest that “typical” TFUS parameters used so far in most studies cause negligible temperature increases in brain tissue [
], this was partly confirmed using the more realistic bio-heat equation to estimate the temperature increase after 40 s of TFUS through a 3 mm thick skull, with an Ispta = 7 W/cm2 in the brain. The maximal increase in the brain was less than 0.2 °C. Interestingly, however, they found rather strong increases in the skull (2.8 °C). Also experimental results show mostly only small temperature increases due to sonication [
]. However, it is important to note that the overall temperature increase depends on the combination of several TFUS parameters. For example, one study [
] reported a measured peak temperature increase of 0.2 °C for an extended stimulation (∼30 min) at a low Ispta≤230 mW/cm2 at 1 MHz (1.6 °C at 5 MHz for otherwise same parameters). In contrast, another study [
] reported a temperature increase up to 3 °C after two blocks of 5 min stimulation at 200 kHz, separated by a 2 min break, at Ispta = 4.5 W/cm2 and a MI = 3.1 (higher than the allowed limit). Both studies applied longer durations than used in most other TFUS studies so far, but the combination with the higher Ispta caused noticeable temperature rises in the second study.
A temperature increase of 0.5 °C was reported through MR thermometry after 30 s sonication at Isppa = 9.9 W/cm2 [
] reported temperature variation within the measurement noise level of the baseline temperatures (±2 °C) with MR thermography. The strongest effect was reported by a study using MR thermometry (sensitivity 0.3 ± 0.06 °C [
], the authors tested the effects of ultrasound stimulation on motor cortex excitability measured by single-pulse transcranial magnetic stimulation (TMS). They report significant changes in the recorded muscle responses to TMS only when it was applied during, but not after, sonication. Follow-up neurological exams and anatomical MRIs after the TFUS experiment did not reveal any abnormalities or changes in the mental or physical status, nor any discomfort associated with the procedure [
] presents results from a follow-up questionnaire after TFUS that could be obtained from 64 out of in total 120 participants. Seven subjects reported mild or moderate symptoms (mild neck pain, scalp tingling, headache, difficulty paying attention, muscle twitches and anxiety) that they felt were possibly or probably related to the experiment. These initial symptoms disappeared upon follow-up. The authors found a linear correlation (r = 0.797, p = 0.0319) between Isppa and the occurrence of observed symptoms among the 7 subjects who reported mild to moderate symptoms that were perceived as ‘possibly’ or ‘probably’ related to participation in TFUS experiments.
Discussion and conclusions
Harmful effects of TFUS were absent in the majority of the 33 studies reviewed here. In two cases, microhemorrhages occurred in a subset of the tested animals when using a high Ispta of 11.2 W/cm2 for a short duration [
]. Both doses are clearly above the safety limits of the FDA guidelines for diagnostic US and above the IEC standard 60601-2-5 for physiotherapy US equipment. However, this also holds for several other included studies, where no adverse effects were reported. While the parameters chosen in one of the studies [
] did not result in substantial heating, as also pointed out by the authors, the high Ispta of 11.2 W/cm2 differentiates it from many other TFUS studies. That might indicate that mechanical effects caused the microhemorrhages, even though the limits for MI and Isppa were not exceeded. However, as this was observed in only one of the tested animals, this conclusion remains very speculative and a replication including sham controls would be favorable to ensure that the microhemorrhages were indeed related to TFUS. In the second study [
], the chosen parameter combination might have led to a high total energy deposit, opening the possibility that a thermal mechanism underlay the adverse effects that occurred in four animals. For example, Gulick et al. [
] observed a temperature increase of 3 °C for a less intense protocol using an Ispta = 4.5 W/cm2 and in total 180 sonications in a time period of 13 min. It seems reasonable to assume that heating might have been even higher in the four animals that showed microhemorrhages in Ref. [
] and Ispta exceeded the FDA limits for diagnostic US in soft tissue in 14 out of the 20 studies in which Ispta was reported or could be calculated post hoc (values after cranial transmission or for craniotomized animals). Ispta was also above the physiotherapeutic limit in 11 of the 20 studies. This suggests that Ispta is the most sensitive safety index in case of TFUS and, unlike current practice, should be reported so that it can be followed up by a more detailed estimation of the thermal effects when its limits are exceeded. We consider this relevant as the current studies indicate that TFUS parameters within the FDA limits for diagnostic US might often lack neural stimulation effectiveness. For example, an Ispta of around 2 W/cm2 for pulsed waves and 4 W/cm2 for continuous waves was necessary to reach a 50% success rate for stimulation at 500 kHz [
] and Table 2). In addition, recent studies show that heating of the skull (potentially causing indirect heating of soft tissue) and/or brain tissue can reach several degrees for more intense and long protocols [
]. The systematic assessment of heating will thus be relevant in future studies that might aim at extending the parameter envelope of TFUS and should be part of any safety test of new sonication regimes in particular for human TFUS.
It is worth noting that the safety limit of 720 mW/cm2 for Ispta, which was generally used in TFUS studies so far and which we also applied here, was introduced to limit the heating in soft tissue. In case of transcranial US, the FDA limits for diagnostic US actually apply an even stricter limit of 94 mW/cm2 for Ispta to prevent excessive heating of the skull, which absorbs most of the beam energy. It seems that almost none of the studies published so far reported neural effects for intensities below this threshold. However, it is important to stress that both limits are based on worst-case scenarios and exceeding them does not necessarily mean that strong heating occurs. Rather, the FDA standard for diagnostic US requires a case-by-case estimation of the maximum temperature rise in soft tissue and skull once they are exceeded, specific for the used ultrasound parameters and setup. Simulations of the propagations of the TFUS beam through the skull, combined with evaluations of the bio-heat equation for TFUS [
], making TFUS a potent neuromodulation modality. This is encouraging for therapeutic applications. In contrast to diagnostic US, future TFUS applications might resort to repeated sessions over extended time periods to achieve and maintain therapeutic efficacy. As such, a safety framework will also need to cover these more intense settings (see, e.g. Refs. [
] for a related example of adverse effects that only occurred after repeated applications in case of transcranial direct current stimulation) or combinations of TFUS with other brain stimulation techniques. This will require safety studies that specifically test this parameter space in order to inform an international consensus on accepted settings and procedures, similar to established non-invasive brain stimulation methods [
Non-invasive electrical and magnetic stimulation of the brain, spinal cord, roots and peripheral nerves: basic principles and procedures for routine clinical and research application. An updated report from an IFCN Committee.
]. This suggests a need for guidelines that provide a secure framework for experimental settings and practical procedures, including mandatory safety screening and appropriate follow-up procedures. For example, the importance of establishing best practices also for apparently simple procedures was highlighted in a recent review [
] of low-intensity low-frequency US (20–100 kHz), showing that US can cause skin damage due to inertial effect cavitation in the coupling gel if non-degassed gel is employed.
Along similar lines, guidelines are important to prevent intensity hotspots that can occur due to unintended standing waves and focusing effects of the skull. While these effects more likely emerge in small animals [
] suggests that unwanted secondary hotspots might have been the cause of intracerebral hemorrhages that occurred in a clinical trial on transcranial low frequency ultrasound for sonothrombolysis [
Transcranial lowfrequency ultrasound-mediated thrombolysis in brain ischemia. Increased risk of hemorrhage with combined ultrasound and tissue plasminogen activator. Results of a phase II clinical trial.
] and that resulted in the early termination of the trial. Finally, the reviewed studies differed in regards to the choice of the stated safety-relevant parameters and the way those were assessed. A more standardized reporting of the relevant pulse parameters and of all safety indices of the FDA guidelines is a prerequisite for the development of future TFUS guidelines for human applications. Accurate estimation of the TFUS intensity after cranial transmission is particularly challenging in humans, as it has to rely on hydrophone measurements based on “representative” skull samples or computer simulations [
], and contributes to variations in the values reported across studies. As such, it seems useful that future studies additionally state intensity values for a pure water background to ensure good comparability of the baseline TFUS parameters.
Conflicts of interest
None declared.
Acknowledgements
We thank the anonymous reviewers for their helpful comments. CP was supported by a PhD stipend of the Technical University of Denmark . HRS holds a professorship in Precision Medicine at the Institute of Clinical Medicine, Faculty of Health and Medical Sciences, Copenhagen University, sponsored by Lundbeckfonden. AT and HRS were supported by a synergy grant from the Novo Nordisk Foundation (Interdisciplinary Synergy Program 2014; grant number NNF14OC0011413 ).
Appendix A. Supplementary data
The following is the Supplementary data to this article:
30338-9&id=gr1.jpg){kind=link}
30338-9&id=gr2.jpg){kind=link}