By name: TCIA CICADAS Skill description: Guide users through creating comprehensive dataset descriptions using the Cancer Imaging Checklist for DAta Sharing (CICADAS).
TCIA CICADAS Skill
Overview
This Skill helps users draft high-quality dataset abstracts and descriptions for TCIA submissions. Following the CICADAS guidelines ensures that datasets are comprehensive, informative, and optimally discoverable.
Documentation Structure
The skill guides the user through the following sections as defined at cancerimagingarchive.net/cicadas:
1. Title
- Full Title: Clear and concise (recommended <= 110 characters).
- Short Title (Nickname): Brief identifier (< 30 characters), alphanumeric and dashes only.
2. Abstract (Maximum 1,000 Characters)
A brief overview including:
- Number of subjects.
- Types of imaging data.
- Types of non-imaging supporting data.
- Potential research applications.
3. Introduction
- Background, purpose, and uniqueness of the dataset.
4. Methods
Collect detailed information across these subsections:
- Subject Inclusion and Exclusion Criteria: Date ranges, demographics, clinical characteristics, and potential study bias.
- Data Acquisition:
- Radiology: Scanner details (vendor, model), parameters (kVp, mA, dose), sequence details (TR, TE, TI, FOV), contrast agents, etc.
- Histopathology: Fixation, staining, scanner info, resolution, and file formats.
- Clinical: Capture process, standards used.
- Other/Missing Data: Genomic/proteomic data and any gaps in completeness.
- Data Analysis:
- File Format Conversions: Software/scripts used (e.g., DICOM to NIfTI).
- Image Preprocessing: Normalization, registration, motion correction.
- Annotation/Segmentation Protocols: Software, guidelines, and observer variability.
- Quality Control: Automated or manual validation steps.
- Automated Analyses: Radiomics, pipelines, and thresholds.
- Scripts, Code, and Software: Specific versions and repository links.
5. Usage Notes
- Data organization and naming conventions.
- Training/test groupings or subsets.
- Instructions for specific file formats.
- Recommended software for viewing data.
- Known sources of error.
6. External Resources
- Links to related datasets, source code (GitHub), or tools stored outside of TCIA.
Conversational Strategy
- Sectional Interview: Don't overwhelm the user. Tackle one section at a time.
- Technical Probing: For the Methods section, ask specific technical questions (e.g., "What was the magnetic field strength for the MRI scans?") based on the acquisition type.
- Refining: Help the user refine their responses into professional, scientific prose.
- Validation: Ensure the Abstract remains under the 1,000-character limit.