TECHNOLOGY

Reading the brain with light.

Fast optical signal measures neuronal activity, not the blood that arrives afterwards. It is non-invasive, it resolves events on the scale of milliseconds, and the hardware is small enough to wear.

What the signal is

When neurons become active, they briefly change shape. Ion and water movement across their membranes makes the cells swell by a small amount, and the surrounding tissue changes with it. Swelling changes how the tissue scatters and absorbs near-infrared light.

We shine light in, measure what comes back, and the scattering change mirrors the electrical event that caused it, leading to a signal 1000x faster than ones that measure blood flow, like fNIRS and fMRI. The signal is called the Fast Optical Signal or Event-Related Optical Signal.

How it differs from fNIRS and EEG

The useful distinction is not how fast each one is — it is what physical quantity each one is actually sensitive to.

What EEG, fNIRS and FOS each measure, and the consequences
Method	What it measures	Timing	Main limitation
EEG	Summed electrical potentials at the scalp	Milliseconds	Low spatial accuracy
fNIRS	Hemoglobin concentration, via absorption	Seconds	Measures blood, not neuronal activity
FOS	Changes in light scattering from neuronal activity	Milliseconds	Low signal-to-noise — the part we're changing

fNIRS and FOS are both optical, but they measure very different things. FOS is an optical measurement of neuronal events.

How the measurement works

A frequency-domain instrument modulates the light source at radio frequency rather than shining a constant beam. When neurons are active, the tissue becomes more transparent and the average path of the photons becomes deeper. This means that on average, photons take longer to reach the detector. With a frequency-domain instrument, we measure this change in photon time of flight as a shift in phase.

Figure 1. The emitted and detected waveforms share a frequency, but phase has shifted versus the reference frequency. The phase shift results from a timing shift.

What is hard about it

SIGNAL-TO-NOISE

The signal-to-noise ratio can be orders of magnitude smaller than that of the hemodynamic signal. Almost every design decision is downstream of that ratio.

BANDWIDTH VERSUS NOISE

In general FOS requires a higher bandwidth than fNIRS hardware. Opening the detection bandwidth wide enough to admit a millisecond optical event also admits more noise. Large surface area detectors often limit bandwidth.

PHASE STABILITY

Phase carries the depth information, so any timing jitter or phase drift over a recording session must be removed.

COUPLING AND MOTION

How well an optode sits against the scalp changes the signal intensity, potentially reducing SNR. Hair, skin tone and movement can all affect the recording. Motion may create artifacts in the signal.

BIOLOGICAL NOISE

Heartbeat and respiration-induced blood flow changes, blood pressure and other factors all modulate the average length of the optical path. They create confounding signals and systemic noise sources not easily reduced by simple frequency response filters.

KNOWING WHEN IT IS REAL

The signal is so small, and historically lots of averaging was used to reduce the noise, but we want single-trial effectiveness.

A note on what we do not publish

We have patent work in flight, so this page describes the public science and the shape of the problem rather than our approach to it. What we can say publicly is that on some of these problems, we have moved further than the published state of the art.

WHERE THIS GOES

Next up

Instrument characterization

Prove the hardware on optical phantoms before trusting it on a person. Done means a scorecard we would show a skeptical reviewer, with a noise floor we can defend and results that repeat across builds.

First human recordings

Record a known, well-replicated response in real subjects — the kind of result the FOS literature already contains, so that agreement means something. Done means matching published effects on our own hardware.

Decoding

Move from detecting that something happened to identifying what. Done means single-trial performance good enough that the measurement is useful in minutes.

We don't publish dates for these.

Safety

The instrument emits near-infrared light at levels well below established exposure limits for skin and tissue. A useful comparison: more photons enter your skull on a sunny afternoon than our instrument delivers. Nothing is implanted, nothing except light penetrates the skin, and a session can be stopped at any moment by taking the headgear off.

Selected references

The FOS literature is public, and reading it is the fastest way to decide whether you believe any of this. Three to start with; the full list, with the sources behind every figure on this site, is on the research page.

Shades of gray matter: noninvasive optical images of human brain responses during visual stimulation.
Gratton, G., Corballis, P. M., Cho, E., Fabiani, M., & Hood, D. C. (1995).
Psychophysiology, 32(5), 505–509.
Fast optical imaging of human brain function.
Gratton, G., & Fabiani, M. (2010).
Frontiers in Human Neuroscience, 4, 52.
The fast optical signal — robust or elusive when non-invasively measured in the human adult?
Steinbrink, J., Kempf, F. C. D., Villringer, A., & Obrig, H. (2005).
NeuroImage, 26(4), 996–1008.