Annotated Capsule Endoscopy Datasets for Clean and Contaminated Regions Segmentation: CECleanliness, Kvasir, SEE-AI, and Corrupted Kvasir

This dataset accompanies the research article titled "Multivariate Gaussian Bayes Classifier with Limited Data for Segmentation of Clean and Contaminated Regions in Small Bowel Capsule Endoscopy Images," submitted to PLOS ONE. The dataset includes four separate files:CECleanliness Database: A collection of original capsule endoscopy images and their corresponding binary ground truth masks, annotated for clean and contaminated regions segmentation in the small bowel.Kvasir Capsule Endoscopy Dataset: A subset of images from the Kvasir dataset, curated for capsule endoscopy studies, with binary masks for segmenting clean and contaminated regions.SEE-AI Project Dataset: Data from the SEE-AI project, containing capsule endoscopy images with expert-annotated ground truth masks, focusing on regions of contamination.Corrupted Kvasir Dataset: A dataset of images derived from the Kvasir capsule endoscopy collection, including corrupted or lower-quality images, alongside their annotated masks.Each image in these datasets is paired with a corresponding binary mask that labels clean and contaminated regions, providing essential data for training and evaluating segmentation models. These datasets were utilized to develop and validate the proposed Gaussian Bayes classifier method for clean-contaminated segmentation in capsule endoscopy images, particularly under limited data conditions.The dataset is intended to support further research in the field of capsule endoscopy, segmentation, and contamination detection, facilitating the development of automated tools for improving clinical outcomes in gastroenterology.