Cervical Cell Images With AgNOR Stain Technique

9,339 images of cervical cells stained with the AgNOR technique, with over 63,000 annotated object instances. Images are sourced from patient slides at the Gynecology and Colonoscopy Outpatient Clinic of the HU-UFSC hospital in Brazil.

Use Cases

• Train an instance segmentation model on over 63,000 annotated object instances to detect cervical cell structures.
• Analyze cellular morphology in 1,600x1200 resolution images with a pixel size of 0.111µm for medical research.
• Develop semantic segmentation models for cervical cell images using expert-generated annotations.

Strengths

• Contains 9,339 high-resolution images (1,600x1200) with a precise pixel size of 0.111µm.
• Features over 63,000 expert-generated annotations for object detection and segmentation tasks.
• Images are sourced from a specific clinical outpatient clinic, providing a focused data context.

Limitations

• The dataset size is categorized as 1K<n<10K, which may be limited for training large-scale deep learning models without augmentation.
• Geographic coverage is limited to patients from a single university hospital in Brazil, which may limit generalizability.
• The specific patient demographics and temporal range of data collection are not detailed.

Provenance

Source

Gynecology and Colonoscopy Outpatient Clinic of the Polydoro Ernani de São Thiago University Hospital (HU-UFSC), Universidade Federal de Santa Catarina.

Collection Method

Images are from 15 different slides stained using the AgNOR technique, with annotations created by experts.

Freshness

Last updated on 2022-07-27.

Geography

Brazil (Region tag indicates 'us', but description specifies a hospital in Santa Catarina, Brazil).