Back to datasets
Dataset assetOpen Source CommunityImage AnalysisCervical Cytology

Cervix93 Cytology Dataset

The dataset contains 93 image stacks and their corresponding extended depth of field (EDF) images, sourced from cases classified according to the Bethesda System as Negative, LSIL, or HSIL. It also includes grade labels for each frame and manually marked points within cervical cells.

Source
github
Created
May 11, 2020
Updated
May 22, 2020
Signals
229 views
Availability
Linked source ready
Overview

Dataset description and usage context

Cervix93 Cytology Dataset Overview

Dataset Description

  • Image Count and Type: 93 image stacks with corresponding extended depth of field (EDF) images, derived from cases graded as Negative, LSIL, or HSIL.

    • Negative: 16
    • Low‑grade Squamous Intra‑epithelial Lesion (LSIL): 46
    • High‑grade Squamous Intra‑epithelial Lesion (HSIL): 31
  • Annotation Information: Grade label for each frame and manually labeled points inside cervical cells.

    • Total manual points: 2,705
    • Negative: 238
    • LSIL: 1,536
    • HSIL: 931

Training and Test Split

  • Training Set (Set 0):

    • Negative: 12 frames, 179 nuclei
    • LSIL: 34 frames, 1,125 nuclei
    • HSIL: 23 frames, 679 nuclei
  • Test Set (Set 1):

    • Negative: 4 frames, 59 nuclei
    • LSIL: 12 frames, 411 nuclei
    • HSIL: 8 frames, 252 nuclei

Code Resources

  • Code Folder: Contains MATLAB evaluation scripts, baseline segmentation methods, and test scripts for the baseline segmentation on the test set.
Need downstream help?

Pair the dataset with AI analysis and content workflows.

Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.

Explore AI studio