Proofread neurons from the MICrONS mouse visual cortex volume, coloured by depth below the pia, gold at the surface through violet to deep blue

microns mouse visual cortexa second dataset

Cortex

The same measured geometry, a different piece of brain. These are proofread neurons from the MICrONS volume, coloured by how deep they sit.

38 cells481 µm of cortex34 located somata

A core through cortex

Every fully proofread neuron I hold from the MICrONS mouse visual cortex volume, 38 of them, seen side on with the pia at the top. Colour is not a property of the cell. It is position: every point on every arbor is tinted by its own depth below the surface, so one pyramidal neuron reads gold where its apical tuft reaches layer 1 and blue where its soma and basal dendrites sit deep. Nothing draws the layers in. They appear because that is where the cell bodies are.

MICrONS cortical neurons coloured continuously by depth below the pia
The bright points are real somata, not estimates. Cell body positions come from the CAVE nucleus table rather than from the meshes, because a mesh centroid is noise inside a soma and the true coordinates are on record. The 34 located cell bodies span 429 to 910 µm below the pia with a median of 630, which is layer 2/3 through layer 6.

the ramp spans

310 – 1083 µm

below the pia. Gold at the top of that range, deep blue at the bottom, violet through the middle. Every point on every arbor is coloured by its own depth, so one neuron carries the whole ramp along its length.

located cell bodies

34 of 38

span 429 to 910 µm, median 630, which is layer 2/3 through layer 6. Four cells have no nucleus in the table and carry no marker.

This uses the same three colour stops as the CA3 convergence gradient, reversed. There the ramp encoded a measured quantity, how many mossy fibers reach a cell. Here it encodes position in tissue. Same visual language, two different kinds of fact, and it is worth saying which is which.

Down through the layers

The same cells, moving. A still of a column is ambiguous in one specific way: you cannot tell whether the cells occupy a slab of tissue or happen to line up from one viewpoint. The only way to answer that is to move the camera.

Ten seconds. The camera swings 72 degrees around the column and descends at the same time, so the apical tufts pass overhead before the deep cells arrive. Colour is still each point's own depth below the pia, so a single neuron reads gold at the top and blue at the bottom no matter which way it is turned. Every cell is at 100 faces per square micrometre of membrane, a density rather than a flat budget, so the large cells carrying those tufts are not the least detailed ones.

What is honest about this picture

These cells were chosen because they are proofread, not because they are a sample of anything. A neuron earns that status by a person checking it, so this is a picture of 38 well traced cells, not of a cortical column. Real cortex at this depth holds tens of thousands of neurons in the same space.

Nine of the mesh identifiers had gone stale since they were downloaded, because proofreading keeps merging and splitting segments, and were resolved forward through the chunked graph before the somata were looked up. Four cells have no nucleus in the table and carry no soma marker.

MICrONS, mouse visual cortex Data from the MICrONS Consortium, served through microns-explorer.org. Meshes and soma positions queried from CAVE, datastack minnie65_phase3_v1. To keep a clip, long press it on a phone or right click it on a desktop.
Back to the CA3 renderings