Galar - a large multi-label video capsule endoscopy dataset

Abstract

Video capsule endoscopy (VCE) is an important technology with many advantages (non-invasive, representation of small bowel), but faces many limitations as well (time-consuming analysis, short battery lifetime, and poor image quality). Artificial intelligence (AI) holds potential to address every one of these challenges, however the progression of machine learning methods is limited by the avaibility of extensive data. We propose Galar, the most comprehensive dataset of VCE to date. Galar consists of 80 videos, culminating in 3,513,539 annotated frames covering functional, anatomical, and pathological aspects and introducing a selection of 29 distinct labels. The multisystem and multicenter VCE data from two centers in Saxony (Germany), was annotated framewise and cross-validated by five annotators. The vast scope of annotation and size of Galar make the dataset a valuable resource for the use of AI models in VCE, thereby facilitating research in diagnostic methods, patient care workflow, and the development of predictive analytics in the field.

Fig. 1

Example images of the 26 labels in the dataset. The figure does not contain images of the labels esophagitis, varices and celiac, as there were no instances of these pathologies present in the set of VCE studies.

Fig. 2

Overall frames per label count of the Galar dataset. Image occurrences per labels are displayed across the three main groups (technical, sections and anatomical). The y-axis is scaled logarithmically. Legend: Orange - Anatomical Green - Pathologies Red - Technical.

Fig. 3

The file structure of the Galar dataset. Frames are stored chronologically in subfolders of the Frames folder. Labels are stored in a single CSV file, per study. The metadata file further contains data on a per study basis.