NeuroHarmonics
TUS Paper Archive
How to citeDownload the datasetGitHub

About

The TUS Paper Archive is a structured database of the primary literature on transcranial and peripheral ultrasound neuromodulation, variously called transcranial ultrasound stimulation (TUS), low-intensity focused ultrasound (LIFU), and transcranial focused ultrasound (tFUS). Each included paper is described by a fixed set of fields covering study design, anatomical target, device, and the acoustic and pulse-timing parameters named in the ITRUSST reporting consensus. The unit of record is the exposure, defined as one anatomical target sonicated at one fundamental frequency.

Scope

A paper is included when it reports an original ultrasound exposure delivered to neural tissue with the intent of modulating its function. All four of the following conditions must hold.

  • Original. The authors performed the exposure themselves.
  • Ultrasound. Acoustic energy above approximately 20 kHz, delivered as a sustained waveform at a carrier frequency.
  • Delivered to neural tissue. The acoustic focus was placed on neurons, nerves, ganglia, plexus, brain, spinal cord, retina or invertebrate nervous system, in culture, slice, in vivo or in a clinical setting. The tissue must be native, not engineered to respond to ultrasound.
  • Intent to modulate function. Not to destroy tissue, open a barrier, deliver an agent, or image.

The following are excluded, each as the negation of one condition.

  • Reviews, meta-analyses, commentaries, editorials, letters and book chapters.
  • Device characterisation, phantom and water-tank measurements, excised-skull and cadaver studies, and computational or simulation studies without an exposure of neural tissue.
  • Ultrasound aimed at an organ whose response is neurally mediated (spleen, gut, heart, kidney, bladder) rather than at a nerve, ganglion or plexus.
  • Thermal ablation including MR-guided focused ultrasound thalamotomy, histotripsy, blood–brain barrier opening, drug and gene delivery, sonothrombolysis, and diagnostic or functional imaging, even when a neural or behavioural outcome is reported.
  • Stimuli that are not ultrasound: transcranial pulse stimulation and other shockwave methods, magneto-acoustic stimulation, mechanical vibration, electrical current, light, audible sound. Ultrasound used only as a power link for an implant or as the imaging modality for another stimulus.
  • Engineered sensitisation: sonogenetics, gas vesicles, nanobubbles, piezoelectric or mechanoluminescent particles, activatable liposomes.
  • Studies whose intended effect is repair rather than a change in activity: regeneration, remyelination, proliferation, differentiation, neurite outgrowth.
  • Conference output of any kind, including abstracts, full proceedings papers, and material without a DOI or equivalent stable identifier.

Inclusion does not depend on the quality of the paper, the robustness of the study design or how completely a paper reports its parameters.

Fields recorded

At the paper level, the following are recorded: model system, species, condition, number of subjects, sessions per subject, randomisation, blinding, sham and control type, auditory confound control, readouts, online or offline measurement, anaesthesia, direction of effect, and adverse events.

For each exposure, the following are recorded: target, coded against a hierarchical vocabulary derived from UBERON; device family, manufacturer and model; fundamental frequency; waveform; pulse duration; pulse repetition frequency; duty cycle; sonication duration; and pressure and intensity in three separate blocks for free-field, in-situ and unspecified domains. Values are never converted between domains or inferred from one block to another. The remaining protocol structure (inter-sonication intervals, trial counts, ramps, nested bursts) is carried as the paper’s own description.

Every quantitative field takes one of three states: a number stated in the paper’s main text, “not reported” where the field applies and the number wasn’t extracted, or “not applicable” where the field does not apply (for example an in-situ estimate for an excised nerve in a bath). A parameter that the paper varied across conditions is recorded as the list of distinct values used.

Methods and verification

Candidate papers are harvested monthly from Europe PMC, PubMed and OpenAlex using a fixed set of neuromodulation phrases, supplemented by citation chaining from consensus documents and systematic reviews. Bibliographic metadata is taken from Crossref. Screening against the inclusion criteria and field extraction from full text are performed by a language model under a fixed prompt. All screening decisions are archived.

Every numeric value is accompanied by the verbatim sentence it was taken from. Duty cycle is recomputed from pulse duration and pulse repetition frequency, and any disagreement with the stated value is displayed. On each paper page, ✓ indicates that the source sentence was verified in the paper text, ✓✓ that an independent second extraction agreed, ? that it disagreed, and ⚑ that the extractor flagged the field for review.

Citation

The archive is released in dated versions of the form YYYY.MM.N. Each release is a git tag in the repository and is identified by the version number shown here and under the Explore table.

NeuroHarmonics (2026). TUS Paper Archive, version 2026.09.0, built 2026-09-24. https://neuroharmonics.com/tus-archive

A methods paper describing the archive is in preparation; this page will carry its reference on publication.

Data access

The complete dataset is available as a single JSON file (see link in header), containing the bibliography, every structured field and every source quotation. Abstracts are displayed on paper pages and indexed for search but are not included in the download. Any filtered selection on the Explore page can be exported as CSV or JSON. The dataset schema is documented in the repository file SCHEMA.md.

Corrections and additions

Each paper page links to a correction form in the GitHub repository with the paper already filled in. The form asks for the field, the value shown, the correct value and the sentence of the paper that states it. To propose a paper the archive does not hold, use the paper proposal form, which asks for the DOI and how the paper meets the four criteria. Corrections are applied as ordinary commits and take effect at the next release.

Maintenance

The archive is built and maintained by NeuroHarmonics. Field names and definitions for the acoustic and timing parameters follow the ITRUSST consensus on standardised reporting for transcranial ultrasound stimulation (Martin et al., Brain Stimulation, 2024). Abstracts are shown with their source recorded.