Definition, intended use, and boundaries among ultrasound technologies
A transcranial ultrasound stimulator uses a transducer to deliver pulsed acoustic energy through the scalp and skull with the aim of modulating a selected neural circuit without creating a permanent lesion. Common research terms include low-intensity transcranial ultrasound stimulation, low-intensity focused ultrasound, and transcranial focused ultrasound neuromodulation. “Low intensity,” “focused,” and “noninvasive” are not complete dose descriptions, and they do not establish that different models share the same focus, mechanism, indication, or safety envelope.
Neuromodulatory stimulation must be separated from other ultrasound uses. Magnetic-resonance-guided high-intensity focused ultrasound can thermally ablate brain tissue; diagnostic ultrasound and transcranial Doppler image anatomy or blood flow; and focused ultrasound combined with intravenous microbubbles can transiently alter blood-brain barrier permeability. Energy, pulse structure, monitoring, risk, and regulatory pathways differ. An approved disease indication for an ablation system, experience with a diagnostic scanner, or findings from a microbubble protocol cannot be transferred to low-intensity neuromodulation.
Transcranial ultrasound is attractive in research because a beam may reach relatively deep subcortical targets within a limited spatial volume. Claims of submillimeter targeting, however, often derive from water-tank measurements or models and do not establish the intracranial focus in every patient. Most neuromodulation uses remain investigational or at an early translational stage. Users must verify the particular product’s authorization, research status, population, and labeling in the relevant jurisdiction rather than infer routine therapeutic approval from a paper, a breakthrough program, research-device status, or a similar appearance.
System components, acoustic parameters, and complete dose
A complete system may include a single-element or phased-array transducer, waveform generator and power amplifier, matching network, coupling membrane or water bag, gel, mechanical holder, optical or electromagnetic navigation, MRI or CT images, acoustic planning software, temperature and output monitoring, and an emergency stop. Wearable research systems may add EEG, EOG, EMG, and state detection. Accessories determine transmission and position. An unvalidated gel pad, water bag, membrane, mount, or replacement transducer changes reflection and focusing and is not an ordinary interchangeable consumable.
A dose report identifies center frequency, number of acoustic cycles, pulse duration, pulse repetition frequency, duty cycle, train duration and interval, pulse count per target, exposure time per session and cumulatively, peak negative pressure, and spatial-peak pulse-average and temporal-average intensity. It states whether values were measured in free field, measured after skull transmission, estimated by a model, or corrected using a specified derating method and provides the assumptions behind mechanical index, thermal index, or thermal-dose calculations. Generator voltage, percentage power, or the phrase low energy cannot reproduce acoustic exposure.
The same temporal-average intensity can be assembled from different peak pressures, pulse widths, and duty cycles with different mechanical and thermal implications. Multitarget, repeated-course, or overnight closed-loop systems must sum all trains and sessions rather than report one pulse. Hydrophone type, scanning medium, calibration date, load, and uncertainty affect confidence in the dose. Calling a free-field value an intracranial dose, or relying only on indices displayed under diagnostic-ultrasound rules, omits skull absorption and sequence-specific effects relevant to neuromodulation.
Skull propagation, target planning, and navigation validation
Skull thickness, density, curvature, sutures, angle of incidence, and adjacent air cavities cause absorption, reflection, phase aberration, focal displacement, side lobes, and local bone heating. A single-element transducer cannot correct individual phase as a phased array can, but both require a defined acoustic window and actual pose. Patient-specific CT more directly informs bone properties, while MRI helps define soft-tissue targets. A template or synthetic CT requires validation of error and population limits, and a colored focus on a planning screen is not an intracranial measurement.
Planning should display the intended and secondary foci, side lobes, skull and skin hot spots, and vessels, ventricles, and sensitive structures along the beam path. Navigation maps transducer coordinates to images but does not prove that transmitted energy remains on target. Registration error, head movement, mount flex, a change in coupling-layer thickness, or an inactive array element can displace the field. Each session should preserve the image and segmentation version, acoustic properties, transducer serial number, pose, contact status, and model output so that deviations can be reconstructed.
Focus size must state whether it is defined from pressure or intensity and at which threshold; axial and lateral dimensions cannot be collapsed into an unexplained diameter. A deep target is not automatically easier or safer. Increasing source output to compensate for transmission loss can increase near-field exposure of skin and bone. A skull defect, cranioplasty, implanted lead, or metal may substantially change propagation and invalidates an intact-skull model unless specifically evaluated. Position changes and turning during overnight use also alter the beam and require loss-of-tracking and coupling-failure stop logic.
Biophysical mechanisms, sensory confounds, and evidence interpretation
Low-intensity ultrasound may influence neural activity through acoustic radiation force, membrane deformation, mechanosensitive ion channels, changes in membrane capacitance, and network-state-dependent effects. Meaningful heating or cavitation is not the therapeutic goal of ordinary neuromodulation. Whether a sequence produces excitation or inhibition cannot be inferred from frequency, duty cycle, or the low-intensity label alone. Cell type, baseline state, pulse structure, and assessment time matter, and findings from slices, anesthetized animals, healthy volunteers, and patients are not automatically interchangeable.
The ultrasound envelope or pulse repetition frequency can generate bone-conducted audible sound, while transducer pressure and skin vibration create somatosensory cues. Auditory startle, altered attention, expectancy, and peripheral nerve co-stimulation can change EEG, motor-evoked potentials, or behavior. Under tighter controls, some earlier findings described as cortical inhibition have been explained by auditory effects. Sham conditions should match sound, touch, device contact, and workflow and should assess whether participants and operators guessed allocation. Masking audio itself can change sleep or EEG and is not a fully inert control.
EEG, functional MRI, evoked potentials, and connectivity changes can support target engagement but do not demonstrate clinical benefit. Concurrent recordings also contain amplifier coupling, transducer vibration, EMG, movement, and filtering artifacts. Studies should prespecify primary endpoints, multiplicity handling, masking, follow-up, and adverse-event capture and should report null findings. A short-term signal at one target neither establishes durable plasticity nor supports treatment of every disease associated with that network.
Participant screening, supervised delivery, and stopping criteria
Before use, a qualified team confirms the investigational or authorized purpose and reviews structural imaging, prior cranial surgery or defects, intracranial lesions or hemorrhage, vascular vulnerability, epilepsy or unexplained loss of consciousness, recent neurologic events, severe headache, auditory and vestibular disease, pregnancy, impaired thermoregulation, and medicines affecting coagulation or seizure threshold. Intracranial electrodes, deep brain stimulators, aneurysm clips, cochlear implants, shunts, cranial plates, and other implants must be identified. Eligibility depends on material, location, beam path, and protocol; being nonmagnetic does not establish ultrasound compatibility.
Recent administration of ultrasound contrast microbubbles or an intent to open the blood-brain barrier creates a different mechanical-risk and monitoring pathway and requires a dedicated protocol. A transducer should not be placed over infected, open, severely inflamed, or poorly coupling skin. Before a session, identity, target, side, image set, and software version are checked, followed by the transducer, array elements, membrane, water bag, gel, and emergency stop. Air is removed from the coupling path and the head is secured. Unknown contact or targeting failure is not corrected by increasing output.
Monitoring includes headache, scalp pain or warmth, nausea, dizziness, tinnitus or hearing change, visual symptoms, numbness, muscle twitching, mood or cognitive change, and any new neurologic sign. Intense or persistent pain, abnormal heating, thunderclap headache, altered awareness, a suspected seizure, focal weakness, language or visual disturbance, persistent auditory or vestibular symptoms, a coupling alarm, loss of tracking, or abnormal generator output requires immediate cessation and assessment. The event record captures the delivered sequence, cumulative exposure, location, coupling, device and accessory lot, concurrent procedures, actions, and outcome.
Closed-loop overnight use, quality control, and governance
A closed-loop system that triggers ultrasound from EEG sleep stage, slow-wave phase, or REM detection requires validation of input quality, algorithm delay, confidence, artifact rejection, repeat-trigger limits, and fallback after loss of lock. Acoustic, motion, or electrical artifacts caused by stimulation must not be classified as the next trigger. Head movement, sweating, patch lift, and gel changes accumulate over a night. If position or coupling cannot be confirmed, the system should stop rather than continue a timer-based plan.
Quality control covers transducer frequency response, element uniformity, output linearity, free-field pressure and intensity, focus and side lobes, pulse timing, harmonics, temperature, coupling detection, navigation accuracy, emergency stop, electrical and electromagnetic compatibility, and software and cybersecurity. Repair, transducer or coupling-component replacement, or changes to an acoustic model, staging algorithm, or firmware require revalidation. Simulation and bench testing should cover worst-case skulls, loads, poses, and cumulative exposure instead of demonstrating performance only under ideal water-tank conditions.
A home-use or wearable form factor does not establish safety for self-directed use. Prescription locking, identity and side checks, session limits, local emergency stop, safe behavior when offline, data integrity, and escalation of adverse events require design validation; a remote log cannot prove the focus remained on target. Services must also govern imaging privacy, raw EEG, algorithm versions, network updates, operator training, calibration traceability, cleaning, and recalls. Publication of a research prototype does not bypass product authorization, ethical review, or long-term risk evaluation.
State of sleep research and clinical role
Transcranial ultrasound can potentially access thalamic, hypothalamic, brainstem, and basal-ganglia sleep-wake circuits, but target engagement is not the same as treating insomnia. A 2026 wearable-patch study compared left subthalamic-nucleus stimulation with a sham night in a fixed order in a small group of young adults and reported a higher REM percentage and shorter REM latency. The sample numbered only a few dozen, the “insomnia” group was questionnaire-screened without a formal diagnosis, sham was always the first night, and sleep efficiency, arousal counts, and other stages did not change significantly. The result is an engineering and physiologic feasibility signal, not proof of efficacy for chronic insomnia, long-term safety, or effective home use.
Sleep studies must especially control the first-night effect, night order, masking noise, transducer sensation, head restraint, and staging bias and should report raw polysomnography quality, blinded human review, next-day function, and longer follow-up. More REM is not inherently beneficial; disease, medication, and psychiatric state change the meaning and risk of REM alteration. Closed-loop animal results or memory changes cannot be directly translated to people, while a protocol for sleep problems associated with tinnitus or another condition is not efficacy evidence before results are published.
Chronic insomnia still requires a standard diagnostic assessment and prioritization of cognitive behavioral therapy for insomnia, which has a more mature evidence base. Obstructive sleep apnea requires assessment of the upper airway and ventilation. Ultrasound stimulation does not splint the airway, provide positive pressure, eliminate respiratory events, or correct hypoxemia, and a change in REM cannot replace respiratory outcomes. At present, sleep-related use should remain within a specific authorized indication or supervised research protocol with prespecified clinical outcomes, stopping rules, and long-term follow-up, without marketing deep targeting as a cure, cognitive enhancement, or diagnosis-free sleep optimization.