SMMILe SpatialAnnotation: Cancer Region Masks for TCGA Whole Slide Images

Pixel-level annotations for cancer regions in TCGA-RCC and TCGA-LU whole slide images, saved as PNG files where white, red, and green mark cancer and blue marks background. Patch-level annotations for TCGA-STAD are provided in CSV files. The dataset was authored by zeyugao and last updated on Hugging Face on 2024-05-15.

Use Cases

• Train semantic segmentation models based on pixel-level cancer region masks.
• Develop patch classification models for cancer detection based on CSV patch annotations.
• Validate image analysis algorithms for histopathology using expert-annotated tissue regions.
• Study spatial patterns of cancer in different tissue types based on the provided annotations.

Strengths

• Annotations are matched to specific TCGA case IDs, providing traceability to source data.
• Annotation mask size corresponds to level 3 of the original whole slide image, ensuring a consistent scale.
• Provides both pixel-level (PNG) and patch-level (CSV) annotation formats for different analysis tasks.

Limitations

• Row count, file formats beyond PNG/CSV, and column-level documentation are unknown, limiting suitability assessment.
• Description metadata is limited; actual data quality and completeness require manual inspection after download.

Provenance

Source

The Cancer Genome Atlas (TCGA) program.

Collection Method

Annotations derived from whole slide images (WSIs) of TCGA-RCC, TCGA-LU, and TCGA-STAD cancer types.

Time Range

null

Freshness

Last updated 2024-05-15 09:29:15; freshness should be verified.

Geography

null