|
|
|
|
|
---
|
|
|
pretty_name: LIDC-IDRI Text-Annotated
|
|
|
license: cc-by-nc-sa-4.0
|
|
|
language: [en]
|
|
|
task_categories: [image-segmentation, image-to-text]
|
|
|
multilinguality: monolingual
|
|
|
source_datasets: [LIDC-IDRI (TCIA)]
|
|
|
size_categories: [10K<n<100K]
|
|
|
---
|
|
|
|
|
|
|
|
|
|
|
|
# 🩺 LIDC-IDRI Text-Annotated
|
|
|
|
|
|
We release a **text-annotated version of the LIDC-IDRI dataset**, where each annotation is **carefully curated from structured metadata** provided by radiologists (e.g., malignancy, size, shape, margin, texture, spiculation, etc.).
|
|
|
|
|
|
This enables new research directions in:
|
|
|
- **Multi-modal learning** (image + text)
|
|
|
- **Text-guided medical image segmentation**
|
|
|
|
|
|
## Includes
|
|
|
- Radiologist Nodule annotations (radiologist contours, malignancy scores)
|
|
|
- Natural language descriptions derived from metadata
|
|
|
|
|
|
> This dataset is based on the publicly available LIDC-IDRI release from TCIA. Text annotations are derived from radiologist-provided metadata.
|
|
|
|
