HCP1065 · 1,065 people

The white matter, and a signal running through it

Eighty seven named cable bundles, averaged over a thousand Human Connectome Project brains, inside that project's own average cortex. The pulses are not decoration: each one is timed by how long its tract really is and how fast you set the signal to travel.

Something a person does

Point of view

Bundle family

Pick one

Conduction velocity

Colour the cortex by

How solid the cortex is

Tractography Yeh FC. Population-based tract-to-region connectome of the human brain. Nature Communications 13, 4933 (2022). doi.org/10.1038/s41467-022-32595-4. HCP1065 atlas, CC-BY-SA 4.0.
Cortical surface Van Essen DC, Smith SM, Barch DM, Behrens TEJ, Yacoub E, Ugurbil K. The WU-Minn Human Connectome Project: an overview. NeuroImage 80, 62 (2013). S1200 group average, fs_LR 32k, MSMAll.
Resting state networks Yeo BTT, Krienen FM, Sepulcre J, et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol 106, 1125 (2011). Parcellation doi.org/10.1152/jn.00338.2011.

Doing something with it

Pick an action above and the scene keeps only the bundles that carry it, in order. The pulse starts at the first leg and each following leg begins where the last one finished, because every leg is offset by the length of everything before it. Reading a word aloud is five bundles and 545 millimetres of cable. Recognising a face and naming it is 554. Telling your other hemisphere something is 131, all of it corpus callosum.

The part worth watching is the arithmetic underneath. At 60 metres per second, reading a word aloud takes about 9 milliseconds of travel, and the five synapses on the way cost another 5. Push the fibres to 120 m/s and the travel halves but the synapses do not move at all, because crossing a synapse takes about a millisecond however fast the cable is. Make the wire fast enough and almost all of the delay is the gaps rather than the wire.

A pathway is a simplification and these are the classical ones. Real behaviour recruits far more than four or five bundles, in parallel, with loops running backwards the whole time. What is honest here is narrower and still worth something: every leg is a real named tract, its length is that tract's own measured median across 1,065 people, and the arithmetic on those lengths is done correctly.

How the two were put in one place

The tracts and the cortex come from different files in different spaces, and lining them up by eye would have been the easiest thing on this page to get wrong and the hardest to notice. They are aligned instead by the affine carried inside the tractography's own file header, which maps a streamline point to MNI coordinates exactly.

The build script then refuses to write anything unless five anatomical facts come out right: the left corticospinal tract has to land on the left, it has to run from the brainstem to the top of the head, the left arcuate has to arch from frontal to posterior temporal on the left, and the corpus callosum has to cross the midline. If the affine were wrong, those fail loudly rather than the page quietly drawing a brain inside out.

Why the animation is a measurement

Every point on every streamline carries how far along its own tract it sits, in millimetres, measured from the real geometry. The shader turns that into an arrival time by dividing by the conduction velocity you choose. So the corticospinal tract, at 133 millimetres, genuinely takes longer to cross than a short cingulum segment, in the correct ratio, and changing the velocity changes every transit time in the scene at once.

This is the one place in the brain where being big does not mean being slow. A metre of axon crossed in a hundredth of a second is what myelin buys, and it is why the impulse sits so far off the diagonal that every other biological process lies on.

What tractography is not. These are not axons. They are the paths a reconstruction algorithm found by following how water diffuses, averaged over 1,065 people. The algorithm can bridge fibres that do not connect and miss ones that do, and nothing in a tractogram says which way a bundle carries signal. What is solid here is that these are the named, replicated bundles of human white matter, in the right places, at the right lengths.
Colours on the cortex. The resting state networks and the 360 region parcellation are the real published label maps, vertex by vertex. The colours assigned to them here are generated to be evenly spaced and distinguishable, not copied from those papers' own palettes, so do not read a colour on this page as the colour that network is usually drawn in.

Another instrument on the same ship

Pain pathways & management. The same cortex and the same measured bundles, turned toward one question: what is pain doing in the nervous system, and at what levels can treatment act? Seven lenses, from nociception to management, with every claim labelled by its kind of evidence and a certainty control that empties the map as the bar rises. And somatotopy paints where movement and touch live on this cortex, area by area.