Artificial intelligence and capsule endoscopy

The advent of artificial intelligence (AI) is revolutionizing today's clinical practice, particularly in gastroenterology, where capsule endoscopy (CE) positions itself as one of the most impacted and benefited areas. Interpreting a video capsule endoscopy procedure, which may involve over 50,000 images per study, is an error-prone, complex process that demands experience and sustained focus. AI, through deep learning models (DL) and convolutional neural networks (CNN), has optimized this reading procedure, demonstrating high sensitivity and specificity in the detection of lesions and providing diagnostic support in pathologies such as Crohn's disease and celiac disease, in addition to significantly reducing reading times. The use of AI as a first-reading tool stands out for its ability to select relevant images, speeding up the review of videos. Tools such as TOP100, already available in clinical practice, have proven effective in reducing reading times, facilitating prioritization in urgent scenarios. However, it faces significant challenges, such as the presence of false negatives. In its role as a second-reading tool, AI can confirm initial findings, detect unseen lesions, and serve as a training tool for physicians in training without compromising diagnostic accuracy. In addition to its impact on reducing reading time, AI has shown potential in detecting intestinal lesions, locating them, and assessing intestinal cleanliness. Despite advances, its widespread adoption requires overcoming significant challenges such as lack of standardization and medico-legal issues. Future prospective, multicenter studies will help AI revolutionize CE interpretation, improving its accuracy and efficiency in clinical practice.