Rhode Island gastroenterology video capsule endoscopy data set

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Small bowel endoscopic evaluation is challenging due to anatomical complexity.
• Video capsule endoscopy (VCE) is the least invasive method for complete small bowel visualization.
• Manual interpretation of VCE data is time-consuming.

Purpose of the Study:

• To present a large, labeled dataset of VCE studies to facilitate the development of AI algorithms.
• To improve the efficiency and accuracy of VCE data interpretation.

Main Methods:

• Compilation of a large VCE dataset from two US hospitals.
• Inclusion of 424 VCE studies and over 5.2 million labeled images.
• Dataset intended for use with existing published data to aid algorithm development.

Main Results:

• A substantial VCE dataset comprising 424 studies and 5,247,588 labeled images was created.
• This dataset, combined with existing resources, can support the development of advanced VCE analysis algorithms.

Conclusions:

• The presented VCE dataset is a valuable resource for advancing AI-driven diagnostic systems in gastroenterology.
• This initiative aims to optimize VCE reading efficiency and improve patient care through automated analysis.