Overview
CALHippo (Cellular Annotation Library for the Hippocampus) is a cell-level dataset for the right human hippocampus in BigBrain, published along the MICCAI 2026 paper (available here) and developed at the University of Modena and Reggio Emilia. CALHippo is designed for cell segmentation, cell classification, and mesoscale hippocampal analysis and is supported by the official CALHippo framework on GitHub.
Why CALHippo
Existing BigBrain-based hippocampal reconstructions are limited by the lack of expert-validated, cell-level annotations across the CA subfields. This limits quantitative modeling, since biologically plausible circuit simulations require reliable estimates of cellular composition and anatomical distribution.
CALHippo addresses this gap by enabling cell-type-aware analysis of the human hippocampal CA complex, supporting research in cell segmentation, classification, density estimation, and quantitative hippocampal modeling.
The Dataset
The release includes:
- High-Resolution Annotations - 24 high-resolution slices with expert-validated annotations across the Cornu Ammonis (CA1, CA2, CA3, and CA4) complexes with class labels (Fig. A).
- Point Cloud Reconstruction - A mesoscale three-class point cloud covering the entire hippocampal CA regions (Fig. B).
Dataset Structure
|
Field |
Value |
|---|---|
|
Region |
Right hippocampal CA complex |
|
Annotated subfields |
CA1, CA2, CA3, CA4 |
|
High-resolution slices |
24 |
|
Resolution |
1 µm/px |
|
Cell classes |
3 |
|
Point cloud |
Included |
|
License |
High-Resolution Annotations
For each CA region and section, the release includes:
*_HR_crop.tif: cropped high-resolution image*_bbox_hr.json: crop bounding box in full-image coordinates*_contours_hr.geojson: CA contour in crop coordinates*_classification_results.geojson: predicted cell contours and classes
Point Cloud
point_cloud.csv contains the columns:
|
Column |
Description |
|---|---|
|
|
Point coordinates in the BigBrain world coordinate system |
|
|
Cell class label: |
|
|
Number of cells sampled at that location |
|
|
CA region label: |
3D Visualization
Point cloud visualization of the three cellular classes in all the right hippocampal CA regions. Points are not single detected cells, as in the csv, but aggregate samples for easier visualization.