Volume electron microscopy of the cerebral microvasculature — precapillary sphincters, pericytes, astrocyte endfeet and the brain's waste-clearance pathway. Open tools and open data for identifying cells in large EM volumes.
build 2026-09-03 08:49
Each card carries a scale model of its volume in the corner — all eight are drawn at the same scale, so their sizes can be compared directly. ωJump holds volumes of very different sizes, so its card stacks five of them, from the smallest on the roster to the largest.
Mouse visual cortex, from the MICrONS minnie65 volume released by the MICrONS Consortium. About 1.7 × 1.4 × 0.5 mm imaged at 4 × 4 × 40 nm. 144,118 nuclei were detected; 94,008 carry a published cell type spanning all sixteen classes, from pyramidal neurons to pericytes, and 50,110 carry none. This is the most heavily annotated of the seven — hand-verified identifications, traced vessel outlines and traced pia and white-matter surfaces from this lab sit on top of the machine predictions.
Mouse visual cortex again, but a different animal and a different volume — the Allen Institute's V1 Deep Dive, spanning pia to white matter. About 1.8 × 1.1 × 0.8 mm at 9 × 9 × 45 nm, coarser than µJump's voxel. 207,455 nuclei were detected, the most of the seven, but its classifier only ever named neurons: 48,077 carry a cell type and 159,378 carry none, so every glial and vascular cell in the block is still waiting to be identified. Cortical layers are fitted from the dataset's own layer-named neurons rather than traced by hand.
Mouse visual cortex from MICrONS Phase 1 — pinky100, the block that came first. About 360 × 215 × 87 µm at the same 4 × 4 × 40 nm as µJump — about a 180th of its volume, which is what the two boxes in the corners of these cards show. 452 somas carry a published call — 359 excitatory, 59 glia, 33 inhibitory — and one is marked uncertain. That is the whole point: µJump's 144,118 nuclei will never be finished, while a roster this size is one a community can actually work through, and the published calls are something to agree or disagree with rather than a table to admire. Synaptic partners come from the proofread soma-to-soma subgraph, embedded in the page.
Mouse visual cortex from the 2016 Lee et al. serial-section TEM volume. About 450 × 450 × 33 µm at 4 × 4 × 40 nm — the same voxel as µJump, and the smallest dataset here. 488 nuclei were detected and none carry a cell type: there is no published segmentation and no classifier output at all, so nothing is pre-named and there are no meshes. It is the only volume here that images the leptomeninges — a continuous sheet of pial and arachnoid cells across the brain surface, with the subarachnoid space above it — and the guided tree's pia/arachnoid branch exists for it. Depth is measured from the traced pial surface rather than fitted.
Human temporal lobe — adult human cortex from the H01 volume (Lichtman lab and Google Research). Roughly a cubic millimetre, a 3 × 2 mm wedge spanning pia to white matter, at 8 × 8 × 33 nm. 47,447 cell bodies were detected; 46,173 carry a published cell type and only 1,274 carry none, making this the most completely classified of the seven. The draw is the species: the same decision tree applied to human tissue, where oligodendrocytes lead every other type more than twofold — 19,352 of them. That is the published classification, not an artefact: just over half sit in the white matter the block runs into. In the cortical layers alone they barely lead the pyramidal cells, 8,761 to 8,089.
The one volume here whose segmentation is not of cells. cb2 is a 998 × 922 × 48 µm block of mouse cerebellar cortex at 4 × 4 × 40 nm, and what it publishes is fragments — pieces of neurite cut wherever the automatic segmentation lost confidence. One Purkinje cell is scattered across dozens of them and no root id gathers them up. So the work here is assembly: click fragments in Neuroglancer, paste the address, and the set becomes a cell, merged into one object by Neuroglancer’s own equivalences — so the link you send somebody shows a cell rather than forty pieces. Not every fragment has a published mesh, and the page counts yours and says so.
Hippocampal CA1 in a diseased brain — a volumetric CLEM dataset of an Alzheimer's model mouse, with amyloid-β, pTau and microglial marker channels registered to the electron microscopy. About 240 × 245 × 26 µm at 8 × 8 × 30 nm. 220 nuclei were detected and none carry a cell type: there is no machine prediction to agree with, so the community is the only classifier. The nuclei were found by blob detection on the Hoechst channel, which makes correcting the detections themselves part of the work.
Not cortex only, and not a classifier. 61 volumes from three sources — twenty Janelia OpenOrganelle FIB-SEM blocks, twenty-eight served through WEBKNOSSOS and thirteen from BossDB, across six species and 29 tissues: heart, kidney, liver, pancreas, thymus, skin, lung, duodenum, cochlea, retina, human airway epithelium and two mechanoreceptor corpuscles alongside cortex, hippocampus, thalamus, nucleus accumbens and a zebra finch song nucleus. The BossDB rows include the only developmental series here — mouse V1 at P6, P14, P105 and P523 against macaque V1 at P7, P75 and P3000 — and immature neuropil is a different identification problem from adult. The FIB-SEM blocks are 4–8 nm isotropic, the finest ultrastructure in this family; the WEBKNOSSOS ones are serial-section, roughly 4–16 nm in plane by 25–50 nm thick, three of them millimetre-scale surveys holding seven tenths of the grid between them. None publishes a soma list or a single cell type, and no detector is being written for them — you are sent to the 40 µm box furthest from anywhere already looked at, 11,707 of them, and reporting an empty box counts for as much as a find.
One account across all eight tools — your points, level and submission history follow you between them. Signing in is optional; you can browse and explore without it.
Søren Grubb is a neuroscientist with a background in neurovascular research and cardiac electrophysiology.
This site hosts research tools built on open, publicly licensed electron microscopy datasets, developed independently as a personal research project.
Cell positions, segmentation and type predictions come from the MICrONS minnie65 mouse visual cortex dataset, released under CC-BY by the MICrONS Consortium (Nature 640, 435–447, 2025). Hand-verified identifications, traced vessel outlines and traced pia/white-matter surfaces are original work from this lab.
δJump uses the Allen Institute's V1 Deep Dive (V1DD) volume of mouse visual cortex, released
publicly as the v1dd_public CAVE datastack
(Allen Institute
Connectomics); viewing its segmentation requires a free CAVE account. Its cortical layer
model is not taken from a published atlas — it is fitted here from the depths of the dataset's
own layer-named neuron types, after correcting for the tilt of the imaged block.
µJump offers an optional Google sign-in. If you sign in, we store your Google display name and email address together with your contribution activity — the cells you identify, IDs you propose, votes, favourites and timestamps — in a private Google Sheet. This is used to attribute contributions, power your personal statistics, and rank a public leaderboard, which shows a chosen display name (never your email). Sign-in is required only to submit or save; you can browse and explore without it. To have your data removed, email soren@grubb.dk.
The seven tools first, then the 61 volumes ωJump carries. Every box is drawn at the same scale — across both tables and the cards above — so any two can be compared directly. The dataset name opens its tool; the publication opens the paper.
| Dataset | Year | Publication | Tissue | Species | Resolution | Size | To scale |
|---|---|---|---|---|---|---|---|
| minnie65µJump | 2025 | MICrONS Consortium, Nature 640 | cortex, V1 L1–6 | mouse | 4 × 4 × 40 nm | 1700 × 1400 × 500 µm 1.2 × 10⁹ µm³ | |
| V1DDδJump | — | Allen Institute — no paper | cortex, V1 full depth | mouse | 9 × 9 × 45 nm | 1800 × 1100 × 800 µm 1.6 × 10⁹ µm³ | |
| pinky100πJump | 2022 | Dorkenwald et al., eLife 11 | cortex, V1 L2/3 | mouse | 4 × 4 × 40 nm | 360 × 215 × 87 µm 6.7 × 10⁶ µm³ | |
| lee16λJump | 2016 | Lee et al., Nature 532 | cortex, V1 L2/3 | mouse | 4 × 4 × 40 nm | 454 × 448 × 33 µm 6.7 × 10⁶ µm³ | |
| H01ηJump | 2024 | Shapson-Coe et al., Science 384 | cortex, temporal full depth | human | 8 × 8 × 33 nm | 3000 × 2000 × 170 µm 1.0 × 10⁹ µm³ | |
| cb2χJump | 2023 | Nguyen, Thomas et al., Nature 613 | cerebellar cortex, lobule V | mouse | 4 × 4 × 40 nm | 998 × 922 × 48 µm 4.4 × 10⁷ µm³ | |
| 3×Tg vCLEMβJump | 2023 | Han, Wei et al., bioRxiv — preprint | hippocampus, CA1 | mouse | 8 × 8 × 30 nm | 240 × 245 × 26 µm 1.5 × 10⁶ µm³ |
One tool, 61 volumes, from Janelia OpenOrganelle, WEBKNOSSOS and BossDB. Grouped by tissue, then species, largest first. None of them publishes a soma list — that is what ωJump is for.
| Dataset | Year | Publication | Tissue | Species | Resolution | Size | To scale |
|---|---|---|---|---|---|---|---|
| Brain · 34 volumes | |||||||
| wk-ppc2 | 2020 | Karimi et al. | cortex L1-5 (PPC-2) | mouse | 11.24×11.24×30 nm | 291 × 622 × 397 µm 7.2 × 10⁷ µm³ | |
| wk-karimi-lpta | 2020 | Karimi et al. | cortex (LPtA) | mouse | 11.24×11.24×30 nm | 391 × 230 × 737 µm 6.6 × 10⁷ µm³ | |
| wk-morgan2020 | 2020 | Morgan et al. | thalamus (dLGN) | mouse | 4×4×30 nm | 444 × 510 × 274 µm 6.2 × 10⁷ µm³ | |
| wk-bock11 | 2011 | Bock et al. | cortex (V1) + leptomeninges | mouse | 4×4×40 nm | 541 × 373 × 49 µm 1.0 × 10⁷ µm³ | |
| wk-bloss18 | 2018 | Bloss et al. | hippocampus | mouse | 3.89×3.89×50 nm | 272 × 393 × 16 µm 1.7 × 10⁶ µm³ | |
| wk-karimi-s1 | 2020 | Karimi et al. | cortex (S1) | mouse | 11.24×11.24×28 nm | 101 × 67 × 219 µm 1.5 × 10⁶ µm³ | |
| wk-gour21-p7-l4 | 2021 | Gour et al. | cortex L4 (barrel, P7) | mouse | 11.24×11.24×30 nm | 77 × 111 × 161 µm 1.4 × 10⁶ µm³ | |
| wk-v2 | 2020 | Karimi et al. | cortex L2/3 (V2) | mouse | 12×12×30 nm | 77 × 98 × 154 µm 1.2 × 10⁶ µm³ | |
| wk-acc | 2020 | Karimi et al. | cortex L2 (ACC) | mouse | 12×12×30 nm | 77 × 146 × 99 µm 1.1 × 10⁶ µm³ | |
| wk-ppc | 2020 | Karimi et al. | cortex (PPC) | mouse | 12×12×30 nm | 78 × 98 × 146 µm 1.1 × 10⁶ µm³ | |
| wk-ex144-l23-s1 | 2022 | Loomba et al. | cortex L2/3 (S1) | mouse | 11.24×11.24×28 nm | 89 × 57 × 214 µm 1.1 × 10⁶ µm³ | |
| bd-wb23-mouse-v1-l4-p105 | 2023 | Wildenberg et al. | cortex L4 (V1, P105) | mouse | 6×6×40 nm | 126 × 139 × 38 µm 6.7 × 10⁵ µm³ | |
| bd-wb23-mouse-v1-l4-p14 | 2023 | Wildenberg et al. | cortex L4 (V1, P14) | mouse | 6×6×40 nm | 90 × 125 × 54 µm 6.0 × 10⁵ µm³ | |
| wk-ex145-l4-s1 | 2019 | Motta et al. | cortex L4 (S1) | mouse | 11.24×11.24×28 nm | 62 × 95 × 93 µm 5.4 × 10⁵ µm³ | |
| bd-wb23-mouse-v1-l4-p6 | 2023 | Wildenberg et al. | cortex L4 (V1, P6) | mouse | 6×6×40 nm | 126 × 137 × 19 µm 3.2 × 10⁵ µm³ | |
| bd-wb23-mouse-v1-l23-p105 | 2023 | Wildenberg et al. | cortex L2/3 (V1, P105) | mouse | 6×6×40 nm | 100 × 108 × 30 µm 3.2 × 10⁵ µm³ | |
| wk-kasthuri2011 | 2015 | Kasthuri et al. | cortex | mouse | 6×6×30 nm | 65 × 80 × 56 µm 2.9 × 10⁵ µm³ | |
| bd-wb23-mouse-v1-l23-p523 | 2023 | Wildenberg et al. | cortex L2/3 (V1, P523) | mouse | 6×6×40 nm | 96 × 112 × 22 µm 2.4 × 10⁵ µm³ | |
| jrc_mus-choroid-plexus-3 | 2021 | Xu et al. | choroid plexus | mouse | 8 nm iso | 60 × 60 × 60 µm 2.2 × 10⁵ µm³ | |
| bd-wb21-nac-dopamine-cocaine | 2021 | Wildenberg et al. et al. | nucleus accumbens (dopamine axons, cocaine) | mouse | 4×4×40 nm | 72 × 97 × 30 µm 2.1 × 10⁵ µm³ | |
| bd-wb23-mouse-v1-l4-p523 | 2023 | Wildenberg et al. | cortex L4 (V1, P523) | mouse | 6×6×40 nm | 83 × 95 × 23 µm 1.8 × 10⁵ µm³ | |
| wk-h5-l23-msem | 2022 | Loomba et al. | cortex L2/3 (STG), MultiSEM | human | 4×4×38 nm | 1409 × 1331 × 88 µm 1.6 × 10⁸ µm³ | |
| wk-h5-l16-msem | 2022 | Loomba et al. | cortex L1-6 (STG), MultiSEM | human | 4×4×37 nm | 1917 × 2553 × 30 µm 1.4 × 10⁸ µm³ | |
| wk-h5-l23-stg | 2022 | Loomba et al. | cortex L2/3 (STG) | human | 11.24×11.24×30 nm | 167 × 216 × 113 µm 4.1 × 10⁶ µm³ | |
| wk-h6-l23-ifg | 2022 | Loomba et al. | cortex L2/3 (IFG) | human | 11.24×11.24×30 nm | 170 × 216 × 79 µm 2.9 × 10⁶ µm³ | |
| wk-mk1-t2-stg | 2022 | Loomba et al. | cortex L2/3 (STG) | macaque | 11.24×11.24×30 nm | 179 × 228 × 108 µm 4.4 × 10⁶ µm³ | |
| wk-mk1-f6-l23 | 2022 | Loomba et al. | cortex L2/3 (S1) | macaque | 11.24×11.24×30 nm | 168 × 220 × 101 µm 3.7 × 10⁶ µm³ | |
| bd-wb23-primate-v1-l23-p75 | 2023 | Wildenberg et al. | cortex L2/3 (V1, P75) | macaque | 6×6×40 nm | 123 × 135 × 38 µm 6.3 × 10⁵ µm³ | |
| bd-wb23-primate-v1-l23-p7 | 2023 | Wildenberg et al. | cortex L2/3 (V1, P7) | macaque | 6×6×40 nm | 124 × 121 × 16 µm 2.4 × 10⁵ µm³ | |
| bd-wb23-primate-v1-l23-p3000 | 2023 | Wildenberg et al. | cortex L2/3 (V1, P3000) | macaque | 6×6×40 nm | 91 × 95 × 27 µm 2.3 × 10⁵ µm³ | |
| bd-wb23-primate-v1-l4-p3000 | 2023 | Wildenberg et al. | cortex L4 (V1, P3000) | macaque | 6×6×40 nm | 125 × 120 × 15 µm 2.2 × 10⁵ µm³ | |
| wk-denk2004 | 2004 | Denk et al. | cortex | rat | 10×10×50 nm | 20 × 18 × 98 µm 3.6 × 10⁴ µm³ | |
| wk-wanner16 | 2016 | Wanner et al. | olfactory bulb | zebrafish | 9×9×25 nm | 121 × 135 × 138 µm 2.2 × 10⁶ µm³ | |
| wk-j0256 | 2017 | Kornfeld et al. | HVC (song nucleus) | zebra finch | 11×11×29 nm | 166 × 166 × 77 µm 2.1 × 10⁶ µm³ | |
| Retina · 5 volumes | |||||||
| wk-k0725 | 2016 | Ding et al. | retina (k0725) | mouse | 13.2×13.2×26 nm | 66 × 211 × 263 µm 3.7 × 10⁶ µm³ | |
| wk-e2006 | 2013 | Helmstaedter et al. | retina (e2006) | mouse | 16.5×16.5×25 nm | 135 × 122 × 86 µm 1.4 × 10⁶ µm³ | |
| wk-k0563 | 2011 | Briggman et al. | retina (k0563) | mouse | 12×12×25 nm | 55 × 66 × 144 µm 5.3 × 10⁵ µm³ | |
| bd-ish21-retina-opl-em2 | 2021 | Ishibashi et al. | retina, outer plexiform layer (EM2) | mouse | 4 nm iso | 27 × 18 × 5.4 µm 2.6 × 10³ µm³ | |
| bd-ish21-retina-opl-em1 | 2021 | Ishibashi et al. | retina, outer plexiform layer (EM1) | mouse | 4 nm iso | 25 × 16 × 4.5 µm 1.8 × 10³ µm³ | |
| Spinal cord · 1 volume | |||||||
| wk-zf-vertebra | 2021 | Guan et al. | spinal cord | zebrafish | 12×12×35 nm | 251 × 188 × 248 µm 1.2 × 10⁷ µm³ | |
| Peripheral nervous system · 3 volumes | |||||||
| jrc_mus-pacinian-corpuscle | 2021 | Xu et al. | Pacinian corpuscle | mouse | 6 nm iso | 58 × 63 × 291 µm 1.1 × 10⁶ µm³ | |
| wk-meissner-sham | 2022 | Gangadharan et al. | Meissner corpuscle (sham, SBEM) | mouse | 11.24×11.24×30 nm | 114 × 72 × 39 µm 3.2 × 10⁵ µm³ | |
| jrc_mus-meissner-corpuscle-1 | 2021 | Xu et al. | Meissner corpuscle | mouse | 6 nm iso | 37 × 46 × 58 µm 9.8 × 10⁴ µm³ | |
| Inner ear · 1 volume | |||||||
| wk-cochlea | 2021 | Hua et al. | cochlea | mouse | 11×11×40 nm | 188 × 101 × 100 µm 1.9 × 10⁶ µm³ | |
| Heart · 2 volumes | |||||||
| jrc_mus-heart-1 | 2021 | Xu et al. | heart | mouse | 8 nm iso | 163 × 152 × 136 µm 3.4 × 10⁶ µm³ | |
| jrc_zf-cardiac-1 | 2021 | Xu et al. | heart | zebrafish | 8 nm iso | 165 × 159 × 321 µm 8.4 × 10⁶ µm³ | |
| Kidney · 3 volumes | |||||||
| jrc_mus-kidney-2 | 2021 | Xu et al. | kidney (kidney-2) | mouse | 8 nm iso | 103 × 101 × 111 µm 1.2 × 10⁶ µm³ | |
| jrc_mus-kidney | 2021 | Xu et al. | kidney | mouse | 8 nm iso | 98 × 64 × 178 µm 1.1 × 10⁶ µm³ | |
| jrc_ut21-1413-003 | 2021 | Xu et al. | kidney (renal carcinoma) | human | 8 nm iso | 121 × 91 × 130 µm 1.4 × 10⁶ µm³ | |
| Liver · 3 volumes | |||||||
| jrc_mus-liver-zon-2 | 2021 | Xu et al. | liver (zonation, zon-2) | mouse | 8 nm iso | 210 × 156 × 427 µm 1.4 × 10⁷ µm³ | |
| jrc_mus-liver-zon-1 | 2021 | Xu et al. | liver (zonation, zon-1) | mouse | 8 nm iso | 189 × 172 × 397 µm 1.3 × 10⁷ µm³ | |
| jrc_mus-liver | 2021 | Xu et al. | liver | mouse | 8 nm iso | 102 × 102 × 71 µm 7.4 × 10⁵ µm³ | |
| Skeletal muscle · 1 volume | |||||||
| jrc_mus-skel-muscle-1 | 2021 | Xu et al. | skeletal muscle | mouse | 8 nm iso | 101 × 61 × 20 µm 1.3 × 10⁵ µm³ | |
| Pancreas · 2 volumes | |||||||
| jrc_mus-pancreas-4 | 2021 | Xu et al. | pancreas | mouse | 8 nm iso | 146 × 113 × 115 µm 1.9 × 10⁶ µm³ | |
| jrc_mus-pancreas-2 | 2021 | Xu et al. | pancreas (pancreas-2) | mouse | 4 nm iso | 20 × 20 × 30 µm 1.2 × 10⁴ µm³ | |
| Thymus · 1 volume | |||||||
| jrc_mus-thymus-1 | 2021 | Xu et al. | thymus | mouse | 8 nm iso | 147 × 116 × 116 µm 2.0 × 10⁶ µm³ | |
| Skin · 2 volumes | |||||||
| jrc_mus-skin-1 | 2021 | Xu et al. | skin | mouse | 8 nm iso | 138 × 114 × 158 µm 2.5 × 10⁶ µm³ | |
| jrc_mus-guard-hair-follicle | 2021 | Xu et al. | guard hair follicle | mouse | 6 nm iso | 88 × 96 × 78 µm 6.6 × 10⁵ µm³ | |
| Airway · 1 volume | |||||||
| jrc_hum-airway-14953vc | 2021 | Xu et al. | airway epithelium | human | 8 nm iso | 213 × 103 × 121 µm 2.6 × 10⁶ µm³ | |
| Lung · 1 volume | |||||||
| jrc_mus-lung-2a | 2021 | Xu et al. | lung | mouse | 8 nm iso | 121 × 62 × 62 µm 4.6 × 10⁵ µm³ | |
| Duodenum · 1 volume | |||||||
| jrc_mus-duodenum-1a | 2021 | Xu et al. | duodenum | mouse | 8 nm iso | 111 × 62 × 41 µm 2.8 × 10⁵ µm³ | |
Sizes are the imaged extent, not the declared array: several of these are published inside a padded address space many times larger, and the difference was measured from the pixels rather than read off the metadata. Together the 68 volumes hold about 4.5 × 10⁹ µm³ of tissue. The smallest box here is under a pixel tall beside H01’s — four orders of magnitude is what that looks like at one scale.