Skip to content
Back to skills

Imaging Data Commons Query

ASecurity

当需要免认证检索/下载 NCI 影像数据公地(IDC)公开癌症影像(CT、MR、PET、病理切片)用于 AI 训练或科研时使用;用 idc-index 按元数据 SQL 查询、批量下载 DICOM、浏览器可视化、核对许可与生成引用;不适用于读写本地 DICOM 像素(用 dicom-medical-imaging)或私有/院内 PACS 数据。触发词:IDC、Imaging Data Commons、idc-index、癌症影像、公开 DICOM、影像数据集、TCGA 影像

  • 3 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 19, 2026
ai-agentspythonbashsqlawsgcpdatabase

Works with

  • cursor
  • cli

Security analysis

A92/100
  • mediumInstalls packages at runtime which could introduce malicious dependencies

Pro shows the line behind each finding and how to fix it

Scanned September 19, 2026

npx -y skills add findscripter/everything-skills --skill imaging-data-commons-query --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Imaging Data Commons Query?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Imaging Data Commons Query
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/findscripter-imaging-data-commons-query/badge)](https://www.skillsdirectory.com/skills/findscripter-imaging-data-commons-query)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: imaging-data-commons-query
title: NCI 影像数据公地查询下载
description: 当需要免认证检索/下载 NCI 影像数据公地(IDC)公开癌症影像(CT、MR、PET、病理切片)用于 AI 训练或科研时使用;用 idc-index 按元数据 SQL 查询、批量下载 DICOM、浏览器可视化、核对许可与生成引用;不适用于读写本地 DICOM 像素(用 dicom-medical-imaging)或私有/院内 PACS 数据。触发词:IDC、Imaging Data Commons、idc-index、癌症影像、公开 DICOM、影像数据集、TCGA 影像
domain: 领域/medical
triggers: [IDC, Imaging Data Commons, idc-index, 癌症影像, 公开DICOM, 影像数据集, TCGA影像, NCI影像]
tags: [idc, idc-index, imaging-data-commons, dicom, cancer-imaging, public-dataset, sql-query, science]
level: 进阶
status: stable
agents: [claude-code, codex, cursor, gemini-cli]
tools: [python, idc-index, pydicom, pandas]
requires: []
related: [dicom-medical-imaging, histolab-wsi-tiling, scientific-database-lookup, pyhealth-clinical-dl]
combines_with: [dicom-medical-imaging, histolab-wsi-tiling, pyhealth-clinical-dl]
license: MIT
source: K-Dense-AI/scientific-agent-skills
source_license: MIT
---
## 何时使用

适用:
- 查找 NCI 影像数据公地(IDC)中的公开放射影像(CT、MR、PET)或数字病理切片(SM)
- 按癌种、模态、解剖部位、厂商等元数据筛选影像子集,构建 AI 训练集
- 从 IDC 云存储批量下载 DICOM(免认证、免出口费)
- 用前核对数据许可(CC BY vs CC BY-NC)并生成署名引用
- 不下载本地,直接在浏览器(OHIF/SLIM)预览序列或整组检查

不该用:
- 读写/匿名化本地已下载的 DICOM 像素与标签 —— 用 `dicom-medical-imaging`
- 私有、院内或非 IDC 来源的影像(IDC 仅含公开数据集)
- 需要全量 DICOM 标签或私有元素的复杂查询 —— 退到 BigQuery(需 GCP 计费账号)

## 步骤

1. 装包并核对版本(**最先做**):`pip install --upgrade idc-index`;用 `IDCClient().get_idc_version()` 确认数据版本(当前 v23),落后则升级。
2. 实例化 `client = IDCClient()`;`index`、`prior_versions_index` 自动加载,其余表需 `client.fetch_index("表名")`。
3. **先探值再过滤**:用 `sql_query` 对 `index` 做 `SELECT DISTINCT Modality/BodyPartExamined ... GROUP BY` 看真实可选值,再带验证过的值正式查询(始终先加 `LIMIT` 试跑)。
4. 选数据:癌种在 `collections_index.CancerTypes`(不在 `index`),需 JOIN;衍生标注/分割用 `analysis_results_index`、`seg_index`、`ann_index`。
5. 下载:`client.download_from_selection(collection_id=... 或 seriesInstanceUID=[...], downloadDir=..., dirTemplate=...)`;大集合分批(每批 10-20 个 series)防超时。
6. 可视化:`client.get_viewer_URL(seriesInstanceUID=...)` 或 `studyInstanceUID=...` 取浏览器链接。
7. 合规:查 `license_short_name`,并用 `client.citations_from_selection(...)` 生成引用(APA/BibTeX/JSON/Turtle)。

## 指令

```python
# 1) 版本核对
from idc_index import IDCClient
client = IDCClient()
print(client.get_idc_version())          # 期望 "v23",落后则 pip install --upgrade idc-index

# 2) 探查可选过滤值(先探后查)
client.sql_query("""
  SELECT DISTINCT Modality, COUNT(*) AS n FROM index GROUP BY Modality ORDER BY n DESC
""")

# 3) 权威列/表结构:client.indices_overview(不必跑 SQL 即可查列与类型)
schema = client.indices_overview["index"]["schema"]
```

## 示例

按癌种 + 模态查(JOIN collections_index):
```python
client.fetch_index("collections_index")
df = client.sql_query("""
  SELECT i.collection_id, i.PatientID, i.SeriesInstanceUID, i.Modality, i.license_short_name
  FROM index i JOIN collections_index c ON i.collection_id = c.collection_id
  WHERE c.CancerTypes LIKE '%Breast%' AND i.Modality = 'MR'
  LIMIT 20
""")
```

下载指定 series(自定义目录层级):
```python
client.download_from_selection(
    seriesInstanceUID=list(df['SeriesInstanceUID'].values),
    downloadDir="./data/breast_mr",
    dirTemplate="%collection_id/%PatientID/%Modality")   # 默认含 %StudyInstanceUID 一级
```

命令行下载(装包后即有,自动识别 collection_id / UID / 清单文件):
```bash
idc download rider_pilot --download-dir ./data
idc download "tcga_luad,tcga_lusc" --download-dir ./data
idc download manifest.txt --download-dir ./data   # 清单每行一个 s3:// URL
```

浏览器可视化 + 许可与引用:
```python
import webbrowser
webbrowser.open(client.get_viewer_URL(seriesInstanceUID=df.iloc[0]['SeriesInstanceUID']))
for c in client.citations_from_selection(collection_id="rider_pilot"):
    print(c)   # 默认 APA;citation_format=IDCClient.CITATION_FORMAT_BIBTEX 出 BibTeX
```

生成清单供复现/分批:
```python
urls = client.sql_query("SELECT series_aws_url FROM index WHERE collection_id='rider_pilot' AND Modality='CT'")
open('ct_manifest.txt','w').write('\n'.join(urls['series_aws_url']))
```

## 注意事项

- 数据分层:IDC 在 DICOM(Patient→Study→Series→Instance)之上加 `collection_id`(原始影像分组,一患者属一集合)与 `analysis_result_id`(跨集合的衍生分割/标注/影像组学)。
- 许可必查:约 97% 为 CC BY(可商用 + 署名),约 3% 为 CC BY-NC(仅非商用),个别自定义条款。发表/商用前务必查 `license_short_name` 并附引用。
- 体量预估:先估大小再下,部分集合达 TB 级;下载文件名是 `<crdc_instance_uuid>.dcm`(非 SOPInstanceUID),云路径 `s3://idc-open-data/<crdc_series_uuid>/<crdc_instance_uuid>.dcm`,可匿名访问、无出口费。
- 列/表以 `client.indices_overview` 为准(随安装版本变化);外部 indices_reference 文档可能超前于本地版本。
- 报错对策:`ModuleNotFoundError` → `pip install --upgrade idc-index`;下载超时 → 减小批量、加重试;查无数据 → 先 `LIMIT 5` 试跑并核对字段名/版本;`BigQuery 配额/计费` → 改用本地 mini-index。
- 仅当 mini-index 与各专用索引(seg/ann/sm/collections)都无所需元数据(如逐段解剖名、SR 定量/定性测量)时,才动用 BigQuery。

## 互见

- related:`dicom-medical-imaging` —— 下载后读写/匿名化/三维重建本地 DICOM
- related:`genomic-file-toolkit`、`scientific-database-lookup` —— 配套科研数据获取
- combines_with:`guided-statistical-analysis`、`single-cell-rnaseq-analysis` —— 下游影像/组学分析流水线

---
本条采编自 K-Dense-AI/scientific-agent-skills(MIT)。

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…