FNIRS · Grey Matter

fNIRS (functional near-infrared spectroscopy) measures brain activity by shining near-infrared light into the head and reading how much returns, and it is the most portable way to follow blood oxygen in the cortex, used in infants, in moving people and outside the lab.

FNIRS. fNIRS (functional near-infrared spectroscopy) measures brain activity by shining near-infrared light into the head and reading how much returns, and it is the most portable way to follow blood oxygen in the cortex, used in infants, in moving people and outside the lab.

Skin, bone and brain are relatively transparent to light between about 700 and 900 nm, and in that window oxygenated and deoxygenated haemoglobin absorb differently. A source on the scalp sends light in; it scatters through the tissue, and a detector a few centimetres away catches the part that has wandered down into the cortex and back. Comparing two or more wavelengths gives the changes in both forms of haemoglobin under the path, and as in fMRI, active cortex shows more oxygenated blood a few seconds after the activity.

It reaches only the surface. The banana-shaped path between source and detector dips about 1.5 to 2 cm, enough to reach the outer cortex under the skull and no further, and the scalp's own blood flow contaminates it.

It is wearable and tolerant. Caps with light sources and detectors run on batteries, are silent, and tolerate movement better than EEG or fMRI, which makes them suited to children, walking, speaking and real settings.

Its resolution is coarse. Diffuse light places activity within a centimetre or more; denser grids of sources and detectors (diffuse optical tomography) improve that substantially.

Like fMRI it reads blood, so it is slow and indirect, which limits it for fast brain-computer interfaces.

Questions: Is a headband that measures blood oxygen in the cortex collecting neural data under the law? The answer is not settled. fNIRS measures haemoglobin in the blood vessels of the cortex, a signal that follows neural activity without being it, while both state laws define neural data as information generated by measuring the activity of the nervous system. California adds that neural data must not be inferred from non-neural information, a clause that a company could read as excluding vascular signals and a regulator could read the other way, since the blood response is used precisely as a measure of neural activity. Colorado's broader category of biological data, which includes physiological properties used for identification, may cover such signals anyway. Why can near-infrared light read only the outer centimetre or two of the brain? Near-infrared light is absorbed little by tissue, but it is scattered every 25 to 200 µm, so after a few millimetres each photon has changed direction many times and carries no image. fNIRS works with that diffuse light: a detector a few centimetres from the source collects photons that wandered down through scalp and skull into the outer cortex and back, and the deepest of them reach only about 1.5 to 2 cm. The method therefore sees the cortical surface under the skull at centimetre resolution, and cannot see deep structures at all.