The readout side of non-invasive BCI. The investable question (Lawrence, non-invasive only): which modality reaches the depth, coverage, spatial resolution and wearable size/weight a real BCI needs, without surgery. The dividing line is depth. Electrical, magnetic and optical modalities are all capped at the cortical surface by the skull; only ultrasound has a physical path to deep structures.
Scorecard (state of the art, Jun 2026)
EEG — depth: cortical surface only, skull-blurred · spatial: poor (~cm) · temporal: excellent (ms) · size: tiny, fully wearable · non-invasive: yes · mature/commodity. A ceiling effect is documented: EEG information content already exceeds what BCI control needs (~175 bit/min), so ML and EEG foundation models (CBraMod, KU Leuven) decode the existing signal better but cannot add depth or spatial resolution. Plays: Neurable, BrainCo, Emotiv, OpenBCI.
fNIRS / HD-DOT — depth: ~20-30mm (cortical) · spatial: HD-DOT reaches ~1/2 fMRI · temporal: slow (hemodynamic, seconds) · size: wearable · non-invasive: yes · maturing. Cannot reach deep structures. Plays: Kernel (Flow), Openwater (optical).
OPM-MEG — depth: cortical (surface-biased) · spatial: better than EEG · temporal: excellent · size: light helmet, BUT needs a magnetically shielded room and often coils, so the environment is bulky · non-invasive: yes · clinically advancing (FieldLine wearable helmet, phase-2 epilepsy/Parkinson’s 2024-25; Cerca 64-sensor; QuSpin). The best wearable surface modality, but shielding-bound and cortical.
fUS (functional ultrasound imaging) — depth: 5-8cm, reaches deep structures · spatial: ~100µm (mesoscopic) · temporal: ~1-10Hz · size: tiny transducer (≈11.5×8.6mm footprint) · non-invasive at depth in adults: NOT YET (current human depth fUSI replaces a skull fragment with the imaging array, i.e. semi-invasive; transcranial adult readout is blocked by skull attenuation and aberration) · research/early but heavily funded. The only modality with the physics for depth + resolution + tiny size; the whole bet is cracking the transcranial skull. Plays: Forest Neurotech (FRO), Merge Labs ($252M, OpenAI/Bain/Newell), Gestala ($21.6M), Openwater.
Note: focused-ultrasound stimulation (TUS, write/neuromodulation) does work transcranially, even to deep targets, because energy only has to go in. fUS imaging (readout) is harder, the echo has to come back out through the skull, which is why transcranial adult imaging lags stimulation.
Verdict
On the depth / coverage / size / weight quartet, fUS is the only non-invasive-class modality that can in principle hit all four; the surface modalities (EEG, fNIRS, OPM-MEG) are structurally capped at the cortex. The gating risk is real and current: non-invasive transcranial adult fUS imaging at depth is unsolved, so today’s deep fUS in humans is semi-invasive (window). The diligence binary for any fUS BCI deal is whether it images at depth through an intact adult skull or needs a window. The smart money (Merge, Forest, Gestala, Openwater) is concentrating on ultrasound precisely to solve this.
How close is transcranial adult fUS (deeper read, 13 Jun 2026)
Three routes through the skull, ranked by invasiveness:
- Cranial window / acoustically-transparent skull replacement — demonstrated in a living adult human (Caltech, Science Translational Medicine; a polymeric window placed during TBI skull reconstruction). Full depth, high resolution, works. Requires the window, so not non-invasive.
- Microbubble-contrast transcranial ULM/fUS — images deep adult human brain vasculature through intact skull and scalp today, but needs an IV microbubble injection plus heavy aberration-correction compute. 4D functional shown in rats through intact skull at ~15µm. No surgery, but contrast-dependent, and whole-brain adult human functional is still preclinical.
- Acoustic-transparency (chelating agent) + pure aberration correction — 2025 preprint: a chelating agent matches skull acoustic impedance for ~94% transmission, ~20µm, full depth, through mouse and human skulls. The cleanest path to truly non-invasive (no contrast, no surgery) if it can be delivered to a living skull safely. Earliest-stage, likely ex-vivo human skull, unproven in vivo.
Route-agnostic enabling layer: skull-aberration-correction compute (ML differentiable beamforming, complex-valued CNNs, SVD, ray-theory), needed by every route, working in rodents and non-human primates and improving fast. The durable wedge that pays off whichever physical route wins.
Revised timeline: a genuinely non-invasive (no-surgery) adult fUS readout at depth is plausibly 2-4 years out, gated by acoustic-transparency translating to living humans, or by contrast-ULM scaling to whole-brain functional.
Who’s on which route (Jun 2026)
The read/write axis explains the field. Transcranial stimulation (energy in) works non-invasively today; transcranial deep readout (echo back out) does not, so the serious readout money goes through a window.
- Forest Neurotech / Merge Labs (Sumner Norman; Forest = Eric Schmidt FRO, Merge = $252M OpenAI/Bain/Newell): fUS imaging readout (plus neuromodulation), but deliberately starting through a skull window / craniectomy because the adult skull is the obstacle (demonstrated across wide swaths of human brain in craniectomy patients, correlations meeting or exceeding fMRI). The best-resourced, most credible readout team, and they chose semi-invasive. That choice is itself the signal that non-invasive transcranial readout is not ready.
- Gestala (China, $21.6M, Guosheng/Dalton): non-invasive transcranial phased-array, but stimulation-led (write, modulating deep circuits), not deep readout. Three months old, prototype pending, building an “Ultrasound Brain Bank” decoding dataset.
- Openwater (Mary Lou Jepsen, ~$100M): non-invasive light + sound + EM imaging, a separate optical/acoustic hybrid.
The combination Lawrence wants, non-invasive + depth + readout, is the hardest corner and is currently unfunded at scale. The only way to back it now is the enabling layer (acoustic-transparency chemistry, aberration-correction compute), or to accept stimulation (Gestala-style) as the non-invasive write play.
Sources: Emergence of fUS for non-invasive BCI (Research/Science) · Future of transcranial ultrasound as a precision brain interface (PMC) · Miniaturized 4D fUS (medRxiv 2025) · OPM-MEG next-gen neuroimaging (Trends in Neurosciences) · Merge Labs $252M ultrasound BCI · Gestala $21.6M (TechCrunch) · HD-DOT depth/resolution (Sci Rep 2025) · EEG foundation models / ceiling