This uses near-infrared light, building on current techniques of using near-infrared light in shorter source/detector paths to measure fluctuations of oxygenated/deoxygenated haemoglobin on the cortex of the brain as a proxy of brain activity. The standard technique is called functional near-infrared spectroscopy, and is similar to fMRI but can only look into superficial brain tissue and not subcortical brain structures like fMRI, with the advantage that it is cheaper, easier, and essentially portable.

In standard fNIRS, a light source and a detector forming a channel have to be ~2-3cm apart. The light leaves the source, goes into the scalp in a banana shape due to refraction, and reaches the detector. The idea is that due to differential absorption of different wavelengths by oxygenated and deoxygenated haemoglobin, you can send 2 wavelengths and solving a 2x2 system gives you the fluctuations in oxygenated and deoxygenated haemoglobin in the tissue the light transversed. This is a proxy of brain activation in that area. If the neurons fire a lot, they consume more oxygen and the brain then sends more oxygen there, this is called Brain-oxygenation level depedent (BOLD) response. If the path length is too short, the light cannot get refracted deep enough to reach the cortex, so you do not measure brain. If it is longer, too much light is absorbed on the way and less signal reaches the detector. The researchers here try to detect light with source/detector diametrically opposite on the scalp, and they show they can. However, it is not clear what kind of application this can have. It was done under very restrictive conditions (subjects very light-skinned, no hair, 30 minutes recording). Moreover, an advantage of standard fNIRS is the high spatial specificity, and it is not clear how to actually translate the light intensity data in their case to brain activation (and probably it is going to be very noisy) as the light transverses all the head.

In any case, they are experimenting with a novel technique, more like a PoC that they can at least detect photons but nothing more than that, and we are probably far away from any potential applications, if any is even come out of this. But it could also lead to applications we cannot actually imagine right now. As for applying this to measure brain activity in the way current fNIRS and fMRI do, I am skeptical.

In case anyone is wondering why (from the paper):

Photons measured in this regime explore regions of the brain currently inaccessible with noninvasive optical brain imaging.

Is this technique useful or already used for other tissues? I'm always surprised how much of my body I can see through with a bright (visible wavelength) light.

Clicked it because I thought they teleported a photon through a head.

Wow, weak signal... "The measured experimental attenuation was found to be of the order 10^18, corresponding to a detection of around one photon per second for a 1.2 W source."

Is this likely to be less risk than xray or nmr mri for comparable imaging quality?

I can't figure out how they would collect depth information from the resulting photon pattern.

I believe reflected photons are much more useful, by measuring how long in between signal and response you can get flight time which tells you depth. Of course I have no idea if infrared light reflects on anything in the brain.

This only works on white patients, so if it actually becomes clinically useful it will be pretty controversial.