Efectividad de la inteligencia artificial en la interpretación de imágenes de cápsula endoscópica para el diagnóstico de enfermedades del intestino delgado: revisión sistemática

Abstract

Introduction: Capsule endoscopic is an effective tool for evaluating the small intestine; however, manual interpretation is time-consuming and prone to errors; therefore, artificial intelligence has emerged as a solution to automate analysis and improve diagnostic accuracy. Objectives: To evaluate the effectiveness of artificial intelligence in interpreting capsule endoscopy images compared to conventional reading for diagnosing small intestine diseases. Methodology - Eligibility Criteria: Original studies published between 2020 and 2025 in English or Spanish that compared AI-based interpretation with conventional reading were included, excluding reviews, case reports, and studies lacking scientific methodology. Databases: Searches were conducted in PubMed, ScienceDirect, and SciELO using DeCS/MeSH descriptors. Risk of bias: National Institutes of Health (NIH) lists were used according to the type of design. Summary of results: The data were organized into descriptive tables, and a narrative summary was conducted according to the type of model and diagnostic performance. Results: Eleven observational and multicenter studies were included, showing that AI applied to capsule endoscopy achieves high sensitivity and specificity, reduces reading time by more than 80%, and outperforms conventional interpretation in detecting intestinal lesions. Limitations: Identified limitations included methodological heterogeneity, publication bias, limited representation of Latin American populations, and non-uniform evaluation criteria. Conclusions: AI increases diagnostic accuracy and reduces capsule endoscopy interpretation time, being an effective complementary tool when used under medical supervision and multicenter validation. Keywords: capsule endoscopy, computer-aided diagnosis, artificial intelligence, small intestine