yangzekang2000/BrainRecon-cubes1937
收藏Hugging Face2026-04-20 更新2026-04-26 收录
下载链接:
https://hf-mirror.com/datasets/yangzekang2000/BrainRecon-cubes1937
下载链接
链接失效反馈官方服务:
资源简介:
---
license: cc-by-4.0
tags:
- neuroscience
- neuron
- brain
- segmentation
- tracing
- reconstruction
- morphology
- microscopy
size_categories:
- 100G<n<1T
configs:
- config_name: default
data_files: annos.json
---
# BrainRecon-cubes1937
A dataset of 1937 brain image cubes with paired neuron morphology annotations for neuron segmentation and reconstruction tasks.
## Dataset Structure
```
BrainRecon-cubes1937/
├── cubes/ # Raw image cubes (.tif)
├── swcs/ # Neuron morphology files (.swc)
├── annos.json # Per-cube annotation metadata
└── data_split/
├── train.txt # 970 samples
├── val.txt # 240 samples
└── test.txt # 694 samples
```
## Data Description
Each sample consists of a **300×300×300 voxel** image cube extracted from a large-scale brain volume, paired with the corresponding neuron morphology in SWC format.
### File Naming Convention
All files share the same stem based on the cube's spatial coordinates in the full brain volume:
```
cube300_x{X}_y{Y}_z{Z}.tif # raw image
cube300_x{X}_y{Y}_z{Z}.swc # neuron morphology
```
### annos.json
A list of 1937 annotation records, one per cube. Each record contains:
| Field | Type | Description |
|---|---|---|
| `id` | int | Unique cube identifier |
| `coord` | list[list[int]] | Bounding box `[[x0,y0,z0],[x1,y1,z1]]` in the full volume |
| `swc_path` | str | Corresponding SWC filename |
| `mean_intensity` | float | Mean voxel intensity of the cube |
| `neuron_ids` | list[str] | Neuron IDs passing through this cube |
| `neuron_length` | float | Total neuron cable length (µm) within the cube |
| `num_nodes` | int | Number of SWC nodes |
| `num_edges` | int | Number of SWC edges |
| `num_fibers` | int | Number of fiber segments |
| `num_branchpoints` | int | Number of branch points |
| `hemisphere` | str | Brain hemisphere: `L` (left), `R` (right), `C` (center) |
| `density` | float | Neuron density metric |
### Dataset Statistics
| Split | Samples |
|---|---|
| Train | 970 |
| Val | 240 |
| Test | 694 |
| **Total** | **1904** |
| Hemisphere | Count |
|---|---|
| Left (L) | 1210 |
| Right (R) | 694 |
| Center (C) | 33 |
| **Total** | **1937** |
| Metric | Min | Mean | Max |
|---|---|---|---|
| Neuron length (voxel) | 388.2 | 1102.7 | 11294.4 |
### data_split format
Each line in `train.txt` / `val.txt` / `test.txt` contains comma-separated relative paths:
```
cubes/<name>.tif,swcs/<name>.swc,mask-r1/<name>_mask.tif
```
## Loading the Dataset
```python
import json
from pathlib import Path
root = Path("path/to/BrainRecon-cubes1937")
annos = json.load(open(root / "annos.json"))
# Load train split
train_files = [
line.strip().split(",")
for line in open(root / "data_split/train.txt")
]
# Each entry: [cube_tif, swc, mask_tif]
```
## License
This dataset is released under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).
许可证:CC BY 4.0
标签:
- 神经科学
- 神经元
- 大脑
- 分割
- 示踪
- 重建
- 形态学
- 显微镜成像
尺寸类别:100G < n < 1T
配置项:
- 配置名称:default
数据文件:annos.json
# BrainRecon-cubes1937
本数据集包含1937个大脑图像立方体,附带配对的神经元形态学标注,可用于神经元分割与重建任务。
## 数据集结构
BrainRecon-cubes1937/
├── cubes/ # 原始图像立方体(.tif格式)
├── swcs/ # 神经元形态学文件(.swc格式)
├── annos.json # 逐立方体标注元数据
└── data_split/
├── train.txt # 970个样本
├── val.txt # 240个样本
└── test.txt # 694个样本
## 数据说明
每个样本均为从大规模大脑体素(voxel)数据中提取的300×300×300体素图像立方体,并配套对应SWC格式(SWC format)的神经元形态学数据。
### 文件命名规则
所有文件均基于其在全脑体积中的空间坐标使用统一的文件名主干,格式如下:
cube300_x{X}_y{Y}_z{Z}.tif # 原始图像文件
cube300_x{X}_y{Y}_z{Z}.swc # 神经元形态学文件
### annos.json
该文件为包含1937条标注记录的列表,每个立方体对应一条记录。每条记录包含以下字段:
| 字段名 | 数据类型 | 说明 |
|---|---|---|
| `id` | int | 唯一立方体标识符 |
| `coord` | list[list[int]] | 全脑体积中的边界框 `[[x0,y0,z0],[x1,y1,z1]]` |
| `swc_path` | str | 对应SWC文件的文件名 |
| `mean_intensity` | float | 该立方体的平均体素强度 |
| `neuron_ids` | list[str] | 穿过该立方体的神经元ID列表 |
| `neuron_length` | float | 该立方体内神经元电缆总长度(单位:µm) |
| `num_nodes` | int | SWC文件的节点总数 |
| `num_edges` | int | SWC文件的边总数 |
| `num_fibers` | int | 纤维段数量 |
| `num_branchpoints` | int | 分支点总数 |
| `hemisphere` | str | 所属大脑半球:`L`(左侧)、`R`(右侧)、`C`(中央) |
| `density` | float | 神经元密度指标 |
### 数据集统计信息
#### 拆分集样本量
| 拆分集 | 样本数量 |
|---|---|
| 训练集 | 970 |
| 验证集 | 240 |
| 测试集 | 694 |
| **总计** | **1904** |
#### 半球分布
| 半球 | 样本数量 |
|---|---|
| 左侧(L) | 1210 |
| 右侧(R) | 694 |
| 中央(C) | 33 |
| **总计** | **1937** |
#### 关键指标统计
| 指标 | 最小值 | 均值 | 最大值 |
|---|---|---|---|
| 神经元长度(体素单位) | 388.2 | 1102.7 | 11294.4 |
### data_split文件格式
`train.txt`/`val.txt`/`test.txt`中的每一行包含逗号分隔的相对路径,格式示例如下:
cubes/<name>.tif,swcs/<name>.swc,mask-r1/<name>_mask.tif
## 数据集加载示例
python
import json
from pathlib import Path
# 数据集根目录路径
root = Path("path/to/BrainRecon-cubes1937")
# 加载标注元数据
annos = json.load(open(root / "annos.json"))
# 加载训练集拆分
train_files = [
line.strip().split(",")
for line in open(root / "data_split/train.txt")
]
# 每个条目格式:[图像立方体tif路径, SWC文件路径, 掩码文件tif路径]
## 许可证
本数据集采用[CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)协议发布。
提供机构:
yangzekang2000



