Radiology Dataset Guide

Radiology Dataset Guide Skill

Triggers

• "radiology dataset"
• "medical imaging data"
• "RSNA challenge"
• "MIMIC data access"
• "CheXpert download"
• "dataset comparison"
• "training data preparation"
• "public dataset"

Parameters

• task_type (required): ML/AI task being solved
• detection - Abnormality/nodule/cancer detection
• segmentation - Organ or lesion segmentation
• classification - Disease or finding classification
• reconstruction - Image reconstruction/enhancement
• quantification - Measurement and feature extraction
• anatomy (optional): Body region or organ system
• modality (optional): Imaging modality preference
• access_requirements (optional): Data use restrictions
• commercial_use (optional): Boolean for commercial application intent

Dataset Inventory

| Dataset | Modality | Primary Task | Access | Annotations | |---------|----------|---------------|--------|-------------| | RSNA Bone Age | X-ray | Regression | Public | Age, quality | | RSNA Pneumonia | Chest X-ray | Detection | Public | Bounding boxes | | RSNA Brain Hemorrhage | CT | Detection | Public | Bounding boxes, type | | NIH ChestX-ray14 | Chest X-ray | Classification | Public | Labels | | CheXpert | Chest X-ray | Classification | Institutional | Labels | | MIMIC-CXR | Chest X-ray | Multi | PhysioNet | Labels, reports | | CheXphoto | Chest X-ray | Classification | Public | Synth/real pairs | | LUNA16 | CT | Detection | Public | Nodule centers | | KiTS | CT | Segmentation | Public | Kidney/tumor | | BraTS | MRI | Segmentation | Research | Multi-modal seg | | PANDA | Histology | Classification | Public | Biopsy grades | | OBJ-CXR | Chest X-ray | Detection | Public | Bounding boxes |

Output Format

Returns structured JSON with:

• Relevant datasets ranked by suitability
• Annotation quality and completeness
• Access procedure and requirements
• Key publications and benchmarks
• Preprocessing recommendations
• Compliance and ethics considerations

Usage Examples

task_type: detection
anatomy: lung
modality: CT

task_type: classification
anatomy: chest
commercial_use: true

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: aizech
• Source: aizech/clinical-skills
• License: MIT
• Homepage: https://www.corpusanalytica.com/skills/

Install and usage instructions live in the source repository linked above.