🔬 Featured Dataset · Cell 2015

Kasthuri11 — Saturated Reconstruction
of Mouse Neocortex

The first fully saturated reconstruction of a volume of mammalian neocortex. All cellular objects and subcellular components annotated at nanometer resolution — hosted and served by the Open Connectome Project.

Access Tokens → API Examples Download Data
6×6×30
nm³ voxel size
1,700+
Annotated Synapses
~40 TB
Raw Data Size
11
Data Tokens
100%
Open Access

Overview

The kasthuri11 dataset is the primary data token for the Kasthuri et al. 2015 Cell paper: "Saturated Reconstruction of a Volume of Neocortex." It represents a landmark achievement in connectomics — the first automated, complete reconstruction of all cellular and subcellular structures within a volume of mouse neocortex at nanometer resolution.

Using automated serial section electron microscopy (ssEM) with tape-collecting ultramicrotomes (ATUM), the Lichtman Lab at Harvard University collected thousands of sections of mouse somatosensory cortex. Every cellular object — axons, dendrites, glia — and subcellular components including synapses, synaptic vesicles, dendritic spines, spine apparati, postsynaptic densities (PSDs), and mitochondria were manually and automatically segmented and stored in an annotated database.

The core scientific finding: by tracing all excitatory axons and recording their proximity to every dendritic spine, the team definitively refuted Peters' rule — the long-held hypothesis that physical proximity between an axon and a dendritic spine is sufficient to predict synaptic connectivity. This online, minable database is hosted by the Open Connectome Project and provides open access to the full dataset.

📋 Dataset Highlights

  • ~1,700 annotated synapses
  • Full axon/dendrite/glia segmentation
  • Synaptic vesicle annotations
  • Dendritic spine & spine apparatus labels
  • Postsynaptic density (PSD) mapping
  • Mitochondria annotations
  • 6×6×30 nm³ resolution (Scale 1)
  • Multiple resolution levels (1–6)
  • Color-corrected EM variant (kasthuri11cc)
  • Refutes Peters' rule experimentally

Methods & Imaging Technology

🔬

ATUM-SEM

Automated Tape-Collecting Ultramicrotome Scanning Electron Microscopy. ATUM collects serial sections continuously onto a tape substrate, enabling long uninterrupted series of sections that cannot be achieved with traditional block-face methods.

⚙️

Automated Segmentation

Computer vision pipelines segment cellular boundaries, classify organelles, and identify synaptic contacts automatically — with expert human proofreading. Segmentation stored in OCP's spatial annotation database.

📐

Volume Specification

The reconstructed cylinder spans the full thickness of a cortical column in mouse somatosensory cortex (S1). Voxel resolution: 6×6×30 nm³ at Scale 1. Multiple downsampled resolution levels available (Scale 1–6).

🧮

Peters' Rule Test

For every excitatory axon, all spatial juxtapositions — both synaptic and non-synaptic — with every dendritic spine were cataloged. The result: proximity alone does not predict connectivity, refuting 50+ years of assumption.

Volume Statistics

Neurons + processes
92% of cellular volume
Glial processes
~8% of cellular volume
Extracellular space
~6% of total volume
Excitatory synapses
~88% of all synapses
Inhibitory synapses
~12% of all synapses
Multi-synapse boutons
documented in dataset

Data Tokens

All tokens are publicly accessible via the OCP RESTful API at openconnecto.me/ocp/ca/<token>/info/. Tokens marked READ-ONLY are raw, unadulterated versions of the data.

Core EM Data

kasthuri11 READ-ONLY Original Electron Microscopy data. The primary raw ssEM image stack at full resolution. info →
kasthuri11cc READ-ONLY Color-corrected version of the EM data. Recommended for most visualization and analysis tasks. info →

Resolution Variants

kasthuri14Maine LOW RES Low resolution dataset for overview visualization and large-scale analysis. info →
kasthuri14s1colEM MEDIUM RES Medium resolution dataset with co-registered color annotations. info →
kasthuri14s1colANNO ANNOTATION Medium resolution annotation layer co-registered with kasthuri14s1colEM. info →

Annotated Regions of Interest

ac3 HIGH RES Small high-resolution annotated region of interest #3. Fully proofread cylinder region. info →
ac4 HIGH RES Small high-resolution annotated region of interest #4. Fully proofread cylinder region. info →

High-Resolution Annotations (Scale 1 · 6×6×30 nm)

kat11segments SEGMENTS Neuron segment labels. Each unique ID corresponds to a single neural process (axon or dendrite). info →
kat11synapses SYNAPSES Synapse annotations (~1,700+). Includes pre/post-synaptic site labels and PSD locations. info →
kat11vesicles VESICLES Synaptic vesicle annotations. Labels individual vesicle clusters within presynaptic boutons. info →
kat11mito MITO Mitochondria annotations across the reconstruction volume. info →

Cylinder Masks (Core Annotation Regions)

kat11mojocylinder MASK Cylindrical mask defining the primary "mojo" annotated core region. info →
kat11redcylinder MASK Cylindrical mask for the red annotated cylinder region. info →
kat11greencylinder MASK Cylindrical mask for the green annotated cylinder region. info →

API Examples

All data is accessible via RESTful HTTP calls. The pattern is: http://openconnecto.me/ocp/ca/<token>/<format>/<resolution>/<x_range>/<y_range>/<z_range>/

# Download 50 MB of original EM data (HDF5)
http://openconnecto.me/ocp/ca/kasthuri11cc/hdf5/1/3000,4000/5000,6000/1000,1050/
# View first slice in browser (XY plane, resolution 1)
http://openconnecto.me/ocp/ca/kasthuri11cc/xy/1/3000,4000/5000,6000/1000/
# Get token metadata / info
http://openconnecto.me/ocp/ca/kasthuri11/info/
# Query synapse annotations in a subvolume
http://openconnecto.me/ocp/ca/kat11synapses/hdf5/1/4400,5424/5440,6464/1100,1200/
# Launch CATMAID viewer for this dataset
http://openconnecto.me/catmaid/?dataview=13

MATLAB CAJAL Toolbox Example

%% Download EM + Annotation data for AC4 region % View online: http://openconnecto.me/ocp/overlay/0.7/ac4/xy/1/4400,5424/5440,6464/1100/ % Install CAJAL: www.github.com/openconnectome/cajal oo = OCP(); oo.setServerLocation('http://openconnecto.me'); oo.setImageToken('kasthuri11cc'); oo.setAnnoToken('ac4'); q = OCPQuery; q.setType(eOCPQueryType.annoDense); q.setCutoutArgs([4400,5424],[5440,6464],[1100,1200],1); anno = oo.query(q); q.setType(eOCPQueryType.imageDense); im = oo.query(q); % Overlay annotations on EM image h = image(im); h.associate(anno);

Python ndio (OCPy) Package Example

# Install: pip install ndio # Or: git clone https://github.com/openconnectome/ndio import ndio.remote.neurodata as nd nd_obj = nd.NeuroData() # Get dataset metadata info = nd_obj.get_token_info('kasthuri11cc') # Download a subvolume of EM data em_data = nd_obj.get_cutout( token='kasthuri11cc', channel='image', x_start=3000, x_stop=4000, y_start=5000, y_stop=6000, z_start=1000, z_stop=1050, resolution=1 ) # Download synapse annotations for the same region synapses = nd_obj.get_cutout( token='kat11synapses', channel='annotation', x_start=3000, x_stop=4000, y_start=5000, y_stop=6000, z_start=1000, z_stop=1050, resolution=1 )

Data Download

Full datasets are available as .tar.gz archives (each z-slice saved as a PNG) or via the Python/CAJAL interfaces above. Total download: hundreds of gigabytes. We strongly recommend using the API to download data in chunks for processing.

📦

Prepared Downloads

Pre-packaged tar.gz archives of commonly requested tokens. Includes kasthuri11, kat11synapses, kat11segments, kat11vesicles, kat11mito.

Download Archive →
🔌

API Chunk Download

Recommended for analysis. Download specific spatial subvolumes in HDF5 format. Use CAJAL or OCPy to automate downloading adjacent chunks for large-scale processing.

❓

Questions?

Post questions or issues to the GitHub issues tracker for the kasthuri2015 repository.

GitHub Repo →

Publication

N. Kasthuri, K. J. Hayworth, D. R. Berger, R. L. Schalek, J. A. Conchello, S. Knowles-Barley, D. Lee, A. Vázquez-Reina, V. Kaynig, T. R. Jones, M. Roberts, J. L. Morgan, J. C. Tapia, H. S. Seung, W. G. Roncal, J. T. Vogelstein, R. Burns, D. L. Sussman, C. E. Priebe, H. Pfister, and J. W. Lichtman. Saturated Reconstruction of a Volume of Neocortex. Cell 162(3):648–661, July 30, 2015.

Cell · 2015 · DOI: 10.1016/j.cell.2015.06.054

Press Coverage

Nature News
Crumb of mouse brain reconstructed in full detail: Digital map is step towards reconstructing a whole human brain
Alison Abbott · Jul 30, 2015
SFARI / Simons Foundation
From tiny slice, researchers lay bare beauty of mouse brain
Jessica Wright · Jul 30, 2015
The Guardian
3D brain map reveals connections between cells in nano-scale
Helen Thomson · Jul 30, 2015
Huffington Post
These Stunning Images Will Take You On A Journey Through The Brain
Carolyn Gregoire · Aug 4, 2015
NIH BRAIN Initiative
BRAIN Investigators Produce Automated Technique Allowing High Resolution Reconstruction of Mouse Cortex
NIH · Aug 11, 2015