Digital Colposcopy Subjective Quality Assessment

Characteristics:

• Multivariate
• Subject Area: Health and Medicine
• Associated Tasks: Classification
• Feature Type: Real
• Instances: 287
• Features: 69 (62 predictive attributes, 7 target variables)
  
  
  Dataset Information:
• Acquired and annotated by professional physicians at 'Hospital Universitario de Caracas'.
• Subjective judgments (target variables) originally done in an ordinal manner (poor, fair, good, excellent) and discretized into two classes (bad, good).
• Images randomly sampled from original colposcopic sequences (videos).
• Original images and manual segmentations included in the 'images' directory.
• Three modalities: Hinselmann, Green, Schiller.
• Target variables: expert::X (X in 0,…,5) and consensus.
  
  
  Features:

1. Image Area:

• cervix_area: Image area with cervix.
• os_area: Image area with external os.
• walls_area: Image area with vaginal walls.
• speculum_area: Image area with the speculum.
• artifacts_area: Image area with artifacts.
• cervix_artifacts_area: Cervix area with artifacts.
• os_artifacts_area: External os area with artifacts.
• walls_artifacts_area: Vaginal walls with artifacts.
• speculum_artifacts_area: Speculum area with artifacts.
• cervix_specularities_area: Cervix area with specular reflections.
• os_specularities_area: External os area with specular reflections.
• walls_specularities_area: Vaginal walls area with specular reflections.
• speculum_specularities_area: Speculum area with specular reflections.
• specularities_area: Total area with specular reflections.
• area_h_max_diff: Maximum area differences between the four cervix quadrants.

1. RGB Color Information:

• rgb_cervix_r_mean, rgb_cervix_g_mean, rgb_cervix_b_mean: Average color information in the cervix (R, G, B channels).
• rgb_cervix_r_std, rgb_cervix_g_std, rgb_cervix_b_std: Stddev color information in the cervix (R, G, B channels).
• rgb_cervix_r_mean_minus_std, rgb_cervix_g_mean_minus_std, rgb_cervix_b_mean_minus_std: (avg - stddev) color information in the cervix (R, G, B channels).
• rgb_cervix_r_mean_plus_std, rgb_cervix_g_mean_plus_std, rgb_cervix_b_mean_plus_std: (avg + stddev) color information in the cervix (R, G, B channels).
• rgb_total_r_mean, rgb_total_g_mean, rgb_total_b_mean: Average color information in the image (R, G, B channels).
• rgb_total_r_std, rgb_total_g_std, rgb_total_b_std: Stddev color information in the image (R, G, B channels).
• rgb_total_r_mean_minus_std, rgb_total_g_mean_minus_std, rgb_total_b_mean_minus_std: (avg - stddev) color information in the image (R, G, B channels).
• rgb_total_r_mean_plus_std, rgb_total_g_mean_plus_std, rgb_total_b_mean_plus_std: (avg + stddev) color information in the image (R, G, B channels).

1. HSV Color Information:

• hsv_cervix_h_mean, hsv_cervix_s_mean, hsv_cervix_v_mean: Average color information in the cervix (H, S, V channels).
• hsv_cervix_h_std, hsv_cervix_s_std, hsv_cervix_v_std: Stddev color information in the cervix (H, S, V channels).
• hsv_total_h_mean, hsv_total_s_mean, hsv_total_v_mean: Average color information in the image (H, S, V channels).
• hsv_total_h_std, hsv_total_s_std, hsv_total_v_std: Stddev color information in the image (H, S, V channels).

1. Fitting Information:

• fit_cervix_hull_rate, fit_cervix_hull_total: Coverage of the cervix convex hull by the cervix.
• fit_cervix_bbox_rate, fit_cervix_bbox_total: Coverage of the cervix bounding box by the cervix.
• fit_circle_rate, fit_circle_total: Coverage of the cervix circle by the cervix.
• fit_ellipse_rate, fit_ellipse_total: Coverage of the cervix ellipse by the cervix.
• fit_ellipse_goodness: Goodness of the ellipse fitting.
• dist_to_center_cervix: Distance between the cervix center and the image center.
• dist_to_center_os: Distance between the cervical os center and the image center.

1. Expert Assessments:

• experts::0 to experts::5: Subjective assessment of each expert (target variables).
• consensus: Subjective assessment of the consensus (target variable).

Citation Of Dataset Papers:

@inproceedings{fernandes2015temporal, title={Temporal segmentation of digital colposcopies}, author={Fernandes, Kelwin and Cardoso, Jaime S and Fernandes, Jessica}, booktitle={Pattern Recognition and Image Analysis: 7th Iberian Conference, IbPRIA 2015, Santiago de Compostela, Spain, June 17-19, 2015, Proceedings 7}, pages={262--271}, year={2015}, organization={Springer} }

@article{fernandes2018automated, title={Automated methods for the decision support of cervical cancer screening using digital colposcopies}, author={Fernandes, Kelwin and Cardoso, Jaime S and Fernandes, Jessica}, journal={Ieee Access}, volume={6}, pages={33910--33927}, year={2018}, publisher={IEEE} }