The use of ultrasound in modern medicine can be first traced to the discovery of piezoelectric phenomena and corresponding piezocrystals by the Curie bothers (Jacques and Pierre) in 1880 [12]. Electrical currents cause physical deformation of these piezo-materials (i.e., transducers), subsequently emitting sound pressure waves in an ultrasound frequency range by tuning the vibration frequency greater than human hearing (> 20 kHz). The utility of ultrasound in biology was explored first in early 1900s for the therapeutic purposes, for example, alleviating muscle/joint pain or indigestion. The ultrasound technology was drastically evolved over the Second World War period [3], in detecting submarines while many of these technologies were later translated to the advent of diagnostic ultrasound imagers and its subsequent and prevailing use in modern medicine. Therapeutic ultrasound had been limited to a few applications [4], such as thermal ablation of the soft tissues, physiotherapies of the muscular conditions, and lithotripsy of the kidney stones. Most of these limitations have been attributed to technical challenges in delivering acoustic pressure waves through bone, whereby the bone introduces significant distortion/scattering/absorption in sound wave propagation [5].

Propagation of the sound waves through any media is complex phenomenon involving linear and non-linear interactions, which go beyond the scope of this review. Therefore, only the basic element of the acoustic physics is introduced. Transmission of the acoustic waves through the media is heavily dependent on their wavelength and corresponding frequencies. Acoustic pressure waves travel in a speed of up to 340 m/s in the air and up to 1,500 m/s in the water-based media while the speed drastically increases in solid media such as bones. The wavelength, as determined by its frequency (Hz) and the speed of the sound waves, is represented by the speed of the sound divided by its frequency. For example, the ultrasound having frequency of 250 kHz in water has a wavelength about 6 mm (i.e., 1,500 m/s ÷ 250,000 Hz). On the other hand, ultrasound used in typical imaging applications, on the order of 2.5 MHz, has therefore, much shorter wavelength of 0.6 mm. Ultrasound typically interacts (through reflection/scattering/absorption) with the objects that are greater than its wavelength, and the ultrasound with a short wavelength (thus high frequency) is therefore used to detect small objects in the body (by detecting the reflective ultrasound ‘echo’ bouncing from the object). On the other hand, to penetrate the bone, for example, skull having nominal 6 mm in thickness, a much longer wavelength (thus lower frequency) ultrasound is needed. For these reasons, a frequency range of 200–700 kHz is used for the transcranial application of therapeutic transcranial ultrasound [6].

FUS technique enables the delivery of highly-focused (with a diameter and length of the focal size measuring less than 10 millimeters) acoustic energy to biological tissue by using an acoustic lens [7], the transducer geometry [8], or the phased actuation of multiple FUS transducer elements [9]. When operating in high-intensity range (typically ranging from a few 100 to 1,000 W), the mechanical energy of the sound waves is transposed to heat which subsequently raises the tissue temperature and causes hyperthermic destruction of the tissue-of-interest. Technical breakthroughs have been made for past decade in transcranial high-intensity FUS methods, whereby phased-array ultrasound configuration (consists of > 500–1,000 small transducer elements surrounding the head) and the independent actuation of transducers are used to correct the acoustic aberration caused by the skull [910]. The technique has been clinically deployed for the thermal ablation of brain tumors and functional neurosurgery in humans [1112]. A single-element FUS configuration, utilizing a segmented spherical shape of the transducer having variable or fixed focal depths, can also deliver focused acoustic energy to the brain transcranially [1314]. In high-intensity ranges, the skull is the major source of acoustic energy absorption; therefore, heating should be carefully monitored. In addition, image-guidance, such as intra-operative magnetic resonance imaging (MRI) and separate magnetic resonance-based navigation systems are used to locate/steer the invisible acoustic focus.

A non-invasive brain stimulation method that enables the controlled modulation of brain activity in a spatially-restricted fashion is sought after as it would offer a new mode of neurotherapeutics. Invasive techniques, such as deep brain stimulation, vagus nerve stimulation, and electrocorticography, are used to modify region-specific cortical function of the brain; however, they carry risks associated with surgery. As a potential alternative to these invasive procedures, transcranial magnetic stimulation (TMS) has been used to modulate cortical activities through the induction of electrical currents on the cortical surfaces by applying strong magnetic fields over the skull [1516]. Transcranial direct current stimulation also allows for non-invasive modulation of brain function by applying weak electrical currents to the scalp and underlying cortical tissues [17]. These modulatory effects have been considered in the context of various clinical applications, including neurorehabilitation and the treatment of major depression [18192021].

Despite the ability to non-invasively modulate brain function, the lack of spatial specificity and penetration depth due to the inductive nature of magnetic stimulation or electrical current conduction severely limits their scope of applications [15]. Optogenetic approaches for brain stimulation [2223], with the ability to control the activity of an individual/group of neural cells, require the genetic modification of neural cells for gaining light-sensitivity, and therefore, are not applicable for immediate translations to humans.

Since a seminal investigation by Fry et al. [24] showing the neuromodulatory potential of FUS in temporarily inhibiting visual evoked potentials by sonicating lateral geniculate nucleus in cats, a series of subsequent studies have found that the administration of FUS, given in a batch of pulses at a low intensity (below the threshold for heat generation or mechanical damage), can reversibly modulate the excitability of regional brain tissue [2526] (detailed review can be found in Bystritsky et al.[27]). We have demonstrated the feasibility of brain stimulation using low-intensity transcranial focused ultrasound (tFUS) through various animal models [282930], and more recently, in humans [14313233]. In addition, tFUS is less prone to generate audible sounds or to create scalp sensations (such as TMS) in humans. Thus, it will be better-suited to provide sham or control conditions during the stimulation, without creating significant cognitive confounders compared to TMS [1314]. As being augmented by the past decade of research around the world, tFUS has gained momentum in showing unprecedented flexibility in depth control as well as spatial resolution of brain stimulation.

The phased-array FUS transducer configuration, typically taking the form of a hemispheric helmet that surrounds the head, is efficient in reaching deep brain areas, as the ‘angle of attack’ of the acoustic wavefront originating from each transducer element is large relative to the skull surface [1234]. However, when the focus is placed near the surface of the brain (for example, cortical areas), the incident angles of acoustic wave front with respect to the skull surface become smaller for a large portion of the transducer arrays, and subsequently increase the level of refraction/reflection at the skull surface beyond control. To remedy this shortcoming, a smaller, single-element FUS transducer can be maneuvered around the skull to deliver FUS to the desired brain location while maintaining the large incident angle to the skull surface (illustrated in Fig. 1). Instead of a helmet-like tFUS device that are filled with degassed water to couple the sound wave between the transducers and the scalp (in the case of a phased-array device), a compressible polyvinyl alcohol hydrogel [35] is provided as a simple form of acoustic coupling media [14313233]. The single-element tFUS configuration does not require surgical-grade preparations for patients (phased-array tFUS typically requires a shaving of the hair and implantation of the neurosurgical stereotactic frame to the skull to fixate the head with respect to the transducer array), which is applicable for practical routine-use in an out-patient setting.

Image-guidance is crucial in applying the small acoustic focus of the sonication to specific anatomies-of-interest, which differ among individuals, both anatomically and functionally. Since the skull may introduce significant distortions in the acoustic propagation, a simple geometric derivation of the acoustic focal location within the cranial cavity is often inaccurate [14]. Thus, the image-guidance also demands an integrated numerical simulation environment, which can estimate the acoustic propagation/focusing within the cranial cavity. The simulation favors the implementation of an expedited processing scheme, e.g., multi-resolution, the finite-difference time-domain formulation to model the transcranial propagation of acoustic waves and the concurrent feedback to the operator [5]. This capability is important since there is no currently-available, non-thermal method to image/track the acoustic focal location in vivo. The neuromodulatory effects take place at a much lower intensity than that can be detected by an MRI technique [36]. The simulation-assisted guidance helps to increase the spatial accuracy of sonication, and may inform the operator of the potential safety risks, such as formation of standing waves [637].

The schematics for single-element FUS transducer configuration is shown in Fig. 1. Typically, 3-dimensional optical registration/tracking system [38], is provided for image-guidance and navigation, whereby the detailed method is described in neuromodulatory studies involving humans [14] and sheep [29]. A structural and functional magnetic resonance imaging (fMRI) are performed to obtain information about the individual's neuroanatomy and to map the relevant brain area. MRI/computed tomography (CT)-visible fiducial markers (e.g., Pinpoint, Beekley Medical, Bristol, CT, USA) are applied to the skin (on skin blemishes, wrinkle lines, and cutaneous veins which offer reliable landmarks spanning over 5 years), serving as reference coordinates for the spatial registration between the image data ‘virtual space’ and the head ‘physical space’ (Fig. 1). An MRI is used to acquire high-resolution T₁/T₂-weighted images from the brain. With the fiducial markers in the same place as the MRI, structural information of the skull is obtained using high-resolution CT, and subsequently spatially co-registered with the fMRI data, to provide the information for sonication planning and real-time guidance, along with on-site simulation-assisted estimation of the location and intensity of the acoustic focus (detailed method is described in Lee et al. [14]). A CT scan also provides information of the calcification that may interfere or absorb the acoustic energy which may impose safety risks, and serves as a screening tool. The image-guidance acquisition process includes the determination of the desired entry/target point, sonication angular orientation, and real-time display of spatial error between the tracked acoustic focus from the target [1438] (Fig. 1).

Sinusoidal ultrasound waves are widely used, and given in a pulsed fashion. The acoustic intensity, i.e., the acoustic power per given area (i.e., W/cm²), is conventionally expressed in spatial-peak pulse-average intensity (I_sppa) while spatial-peak temporal-average intensity (I_spta) represents its time-averaged value per each stimulus. FUS is given in a batch of pulsed sinusoidal pressure waves at a fundamental frequency of 250 kHz, typically within 200–700 kHz range for transcranial application [6]. The individual pulses, each having a specific tone-burst-duration (TBD), are administered repeatedly at a certain pulse repetition frequency (PRF) (Fig. 2 for the schematics), whereby TBD and PRF together determine the duty cycle of sonication (expressed in %, indicating the fraction of active sonication time per given quanta). The overall duration of pulsed sonication is described as the sonication duration (SD). We have found that stimulatory effects are elicited in high duty cycles (50%–70% range) at short SD, whereas suppressive effects are seen at lower duty cycles (5% or less) and longer SD (few tens of seconds) [283039]. The jury is still out about exact parameters and whether the effects are occurring at single cell-level, and further study is necessary to reveal the exact parameters and underlying mechanisms (Fig. 2).

Although the detailed mechanism behind this ultrasound-mediated neuromodulation has yet to be ascertained, transient deformation of the neuronal cell membrane, and associated changes in trans-membrane capacitance/subsequent modulation of action potentials [404142], or the manipulation of cellular excitability of glial cells mediated via mechanoreceptors [43], have been suggested as a few of the likely candidates. Recent investigations on rodent models have also suggested the potential involvement of auditory areas in mediating physical response to FUS [44], in forms of auditory startle responses. Although these seemingly confounding effects have yet to be seen in humans and non-human primates, one should be careful in designing experiments with consideration of potential involvement of the neural substrates other than the ones that are directly stimulated by FUS.

One of important questions in evaluating the utility of tFUS in neurorehabilitation is whether the effects of sonication can actually outlast the SD itself. In order to have therapeutic effects to take place, one should examine the presence of long-term effects of sonication, enough for inducing the neural plasticity. We transcranially applied FUS to the somatosensory areas of the anesthetized rats for 10 minutes at a low duty cycle (5%) and intensity, and measured the time-progression of somatosensory evoked potentials (SEPs) induced by the unilateral electrical stimulation of the hind limb [45]. Compared to the control sham condition (which did not involve any sonication), differential SEP features were observed and found to persist beyond 35 minutes after stopping the administration of FUS. Although there are possible confounders due to the use of anesthesia (ketamine and xylazine) in this rodent study, the presence of this non-transient neuromodulatory effect provides early evidence that FUS-mediated brain stimulation has the potential to induce neuroplasticity. The utility of long-term, repeated administrations of the technique and its clinical efficacy in setting in more permanent neuroplasticity, especially for the context of neurorehabilitation, calls for further investigation.

One should pay close attention to sonication parameters so that they will not exceed major safety-guidelines, i.e., Mechanical Index < 1.9, I_sppa < 12 W/cm², and I_spta ≤ 3 W/cm² (European standard for ultrasound stimulation device [46]) to maximize the safety [47]. Particularly relevant for the transcranial application, potential pressure reverberation [48] (resulting in inhomogeneous intensity profile near the focus) and the formation of standing waves (pressure waves outside of the intended focus, caused by the undesired positive interferences within the cranium cavity) [49], should be avoided. It is also important to monitor the presence of potential skull heating in the ultrasound beam path if the FUS is given relatively close to the skull base (the thalamus is one of the examples). Also, since a stroke or cerebral hemorrhaging may impact the mechanical integrity of the affected brain regions, making them particularly vulnerable to any external mechanical agitation (which is caused by the FUS), additional restrictions and cautions in performing tFUS are advised. Due to these unknown risks involving the patient population, tFUS has been conceived and applied to only a few clinical arenas (e.g., Alzheimer's disease, epilepsy, and modulation of consciousness in coma/vegetative state/minimally-consciousness state).

The past decade of research has shown that the pulsed application of FUS, when combined with the intravenous injection of microbubbles (clinically used in ultrasound imaging), can temporarily and reversibly disrupt the regional blood-brain barrier (BBB) detailed review is provided in Cammalleri et al. [50]. This technique has shown new possibility for transferring exogenous pharmaceutical/biological agents, including genetic materials and even cells across the BBB. These new possibilities will provide a niche for delivering therapeutic agents for neurorehabilitation.

Other than applications in neurorehabilitation, tFUS brain stimulation may open new non-invasive therapeutic avenues for the treatment of neurological/psychiatric conditions. For example, tFUS may be applied to modify aberrant prefrontal cortex function associated with drug-resistant depression [51]. Also, modulation of the reward neural substrates, such as ventral striatum [5253], may someday help to treat substance abuse. The ability to modulate the function of spatially-selective brain areas may also provide the technical foundations for non-invasive assessment/mapping of brain dysfunction, for example, an interrogation of normal or aberrant neural activity. We also note that identical tFUS techniques can be used to effect the neural conduction through myelinated nerves and to elicit tactile sensations by directly stimulating the nerve endings [5455], suggesting its further applicability in the peripheral nervous system, with potential utility in modulating white matter tracts in the brain.