Cervical Cytology Images with Pathological Descriptions

Singpath-CytoText is a cervical cytology image dataset with structured pathological descriptions and annotations. Each image includes a bounding box for an anchor cell, a structured description of nuclear features, a natural language caption, and a final cytological label. The dataset was created by zqiu96 and was last updated on the Hugging Face platform in February 2026.

Use Cases

• Train object detection models to locate anchor cells using the provided bounding box annotations.
• Develop vision-language models to generate natural language captions from cytology images and structured features.
• Classify cytological labels (e.g., ASC-H) from image data and structured descriptions like nuclear size and chromatin pattern.
• Analyze correlations between structured pathological features (nuclear membrane status, chromatin pattern) and diagnostic outcomes.

Strengths

• Dataset includes multiple annotation types per image: bounding boxes, structured descriptions, captions, and labels.
• Data is hosted on a major platform (Hugging Face) with a documented last update in 2026.

Limitations

• Specific row count, image resolution, and dataset size are unknown.
• Potential class imbalance or geographic bias is not described in the provided information.

Provenance

Source

Hugging Face dataset repository created by user zqiu96, associated with the Singpath-VL paper.

Collection Method

null

Time Range

null

Freshness

Last updated on the platform in February 2026.

Geography

null