Dataset · 2025
MICrONS — mouse visual cortex
A cubic millimetre of mouse visual cortex where the same neurons were both recorded in a living animal and reconstructed in EM. It is the only large volume that lets you ask what a measured cell is wired to.
| Species | Mouse |
|---|---|
| Region | Visual cortex (VISp, VISrl, VISal, VISlm) |
| Volume | 1.4 × 0.87 × 0.84 mm |
| Cells | ~200,000 anatomical; ~75,000 functionally imaged |
| Synapses | 523 million |
| Data size | 1.6 PB |
| Resolution | 4 nm in-plane, 40 nm sections |
| Modality | Serial-section EM co-registered with two-photon calcium imaging |
| Access | Open; CAVEclient for programmatic access |
| Status | Released |
MICrONS is the dataset behind almost every structure-and-function question on this site. An awake mouse viewed natural and synthetic stimuli while calcium imaging recorded around 75,000 neurons; the same tissue was then imaged by EM and reconstructed, yielding roughly 200,000 cells and 523 million synapses. Co-registration links a subset of the functional cells to their anatomical reconstructions.
What it is good for. Any question of the form “does the wiring predict the response?” — and for learning proofreading against a segmentation that is good but demonstrably not finished. The CAVE infrastructure behind it is also the practical introduction to versioned connectomics data: every query you run must pin a materialization version or it is not reproducible.
What it does not support. Treating the functional and anatomical populations as the same set. Co-registration succeeds for a fraction of cells, and that fraction is not random — it is biased toward cells near the imaged surface. Synapse counts from the automated pipeline carry both false positives and misses; the published analyses correct for this and yours must too.
Where these numbers come from. MICrONS Consortium (2025), Nature 10.1038/s41586-025-08790-w; microns-explorer.org. DOI 10.1038/s41586-025-08790-w.