CIC_Morph / README.md
ybb2297's picture
Update README
7805cd6 verified
|
Raw History Blame Contribute Delete
6.26 kB
metadata
license: cc-by-nc-4.0
pretty_name: Hippocampal formation single-neuron reconstructions, compartment-labeled (SWC)
tags:
  - neuroscience
  - neuron-morphology
  - swc
  - mouse
  - hippocampus
  - allen-ccf
size_categories:
  - 1K<n<10K
configs:
  - config_name: metadata
    data_files: metadata.csv
    default: true
  - config_name: saustad2026_SUB
    data_files: saustad2026_SUB/metadata.csv

Hippocampal formation single-neuron reconstructions, compartment-labeled

8,321 mouse single-neuron reconstructions from the hippocampal formation (CA1, CA2, CA3, DG, PARA, POST, PRE), registered to the Allen CCFv3. Every node has been labeled by compartment: soma, axon, basal dendrite or apical dendrite.

A separate section, saustad2026_SUB/, adds the 689 subicular reconstructions from Saustad et al. (2026). Those files keep the authors' original node types, and their labels are the authors' column, layer and class (see below).

Quick start

One region, no code: download the zip you need from zips/, for example zips/CA1_full_morph.zip. Each zip holds the SWCs, their change logs and a metadata.csv for that group.

Python:

from huggingface_hub import snapshot_download
import pandas as pd

repo = "ybb2297/CIC_Morph"
meta = pd.read_csv(f"hf://datasets/{repo}/metadata.csv")

# only CA1 full morphologies
path = snapshot_download(repo, repo_type="dataset", allow_patterns=["swc/CA1/full_morph/*"])

Layout

metadata.csv                        one row per neuron (start here)
summary_by_region.csv               counts per region / type
swc/<REGION>/<type>/<id>.swc        labeled reconstructions
change_logs/<REGION>/<type>/<id>_change_log.csv
zips/<REGION>_<type>.zip            same files, one archive per group
zips/saustad2026_SUB.zip
saustad2026_SUB/
  metadata.csv                      authors' column / layer / class / targets per neuron
  swc/<column>/<id>.swc

<type> is full_morph or projections:

  • full_morph: complete morphologies (axon and dendrites), labeled by hand.
  • projections: projection-only reconstructions. The root is the soma (type 1) and every other node is axon (type 2).

Contents

Neurons per region. Every cell in the labeled set is compartment-labeled; "QC ok" means no unlabeled nodes remain (see qc_flag).

Region Full morphology Projection-only Total neurons With change log QC ok
CA1 1,477 2,216 3,693 3,551 3,043
CA2 47 — 47 45 38
CA3 471 425 896 836 748
Dentate gyrus (DG) 1,326 1,733 3,059 2,169 2,594
Parasubiculum (PARA) 129 — 129 101 88
Postsubiculum (POST) 168 — 168 116 113
Presubiculum (PRE) 329 — 329 221 204
Labeled set total 3,947 4,374 8,321 7,039 6,828
Subiculum (SUB), Saustad et al. 689 — 689 — —
All neurons 4,636 4,374 9,010

Saustad et al. subicular neurons by the authors' column and class:

Column Neurons IT ET CT
SUBdd 167 137 0 30
SUBdv 83 64 0 19
ProSUB 111 0 59 52
SUBv 273 40 138 95
SUBvv 55 33 12 10
Total 689 274 209 206

File format

Standard SWC, with seven whitespace-separated columns: id type x y z radius parent. Coordinates are Allen CCFv3 micrometres in axis order AP, DV, ML. The root has parent -1.

type meaning
1 soma
2 axon
3 basal dendrite
4 apical dendrite
0, 10–12 unlabeled (only in cells flagged below, and in the Saustad section)

Change logs record every type change made during labeling, relative to the original reconstruction. Columns: node_id, old_type, new_type, scope, root_id. The 1,258 older logs have no root_id column. A cell has no log when labeling did not change any node type.

metadata.csv

column meaning
neuron_id brain_cell id, e.g. 202254_016
region, morph_type folder the cell sits in
swc_path, change_log_path relative paths (empty if no log)
n_nodes, n_soma, n_axon, n_basal, n_apical, n_unlabeled node counts per type
n_roots number of root nodes. Cells with more than one root are reconstructions with separate fragments (multi_fragment, 55 cells)
n_type_changes nodes whose type differs from the original reconstruction
source_batch, source_set labeling batch and set the cell came from
qc_flag ok (6,828), minor_unlabeled (1,485: a few nodes left unlabeled, almost always well under 1% of the cell), needs_relabel (8: labeling incomplete; exclude them for compartment-level analyses)

Saustad et al. subicular set (saustad2026_SUB/)

689 subicular neurons from Saustad et al., Mapping Projectome Heterogeneity of Subiculum Neuron Cell Types (Hippocampus, 2026). The files are in the same CCFv3 frame and axis order as the main set. The two sets share no neuron ids.

  • Node types are the authors' originals and are not compartment-labeled. About 69% of nodes are type 0, 28% axon and 4% basal dendrite.
  • metadata.csv columns:
    • column: the authors' subdivision. SUBdd 167, SUBdv 83, ProSUB 111, SUBv 273, SUBvv 55.
    • sub_layer: the authors' layer, 1–4.
    • cls: class derived from the layer. Layers 1–2 are IT (274), layer 3 is ET (209), layer 4 is CT (206).
    • targets, n_targets: the authors' projection targets.
    • Node counts per type, as in the main metadata.
  • Membership and labels come from the Neuron_IDs sheet of the authors' spreadsheet. Nothing has been relabeled.

Paper: https://onlinelibrary.wiley.com/doi/10.1002/hipo.70138 (doi: 10.1002/hipo.70138)

Please cite Saustad et al. (2026) when you use this section, and see the paper for methods, definitions and data availability. The license terms of the original authors apply.

License

CC BY-NC 4.0 for the labeled set. The Saustad et al. section is redistributed under its authors' terms.