# Artificial intelligence in advanced endoscopic imaging: transforming optical diagnosis in gastroenterology

## TL;DR

Artificial intelligence is improving endoscopic imaging in gastroenterology by enhancing diagnosis and reducing unnecessary procedures.

## Contribution

The paper highlights AI's novel role in real-time lesion detection and decision-making in both upper and lower GI endoscopy.

## Key findings

- AI improves detection of dysplasia in Barrett’s esophagus and early gastric cancer.
- AI-assisted colonoscopy increases adenoma detection rates and reduces interval colorectal cancer.
- Multimodal AI approaches combine imaging with clinical data to enhance diagnostic precision.

## Abstract

The term Artificial intelligence (AI) is revolutionizing gastrointestinal (GI) endoscopy by enhancing advanced imaging techniques such as Narrow Band Imaging (NBI), Linked Color Imaging (LCI), iSCAN, and Confocal Laser Endomicroscopy (CLE). AI-driven deep learning algorithms, particularly convolutional neural networks (CNNs) and transformer-based models, have demonstrated high accuracy in the real-time detection, classification, and risk stratification of premalignant and malignant lesions, thereby reducing unnecessary biopsies and improving diagnostic efficiency. In the upper GI tract, AI has shown superior performance in detecting dysplasia in Barrett’s esophagus, distinguishing early gastric cancer from benign alterations, and predicting submucosal invasion depth. This capability enhances decision-making regarding endoscopic resection, such as endoscopic submucosal dissection (ESD). In the lower GI tract, AI is increasingly applied for real-time identification of adenomas, serrated lesions, and neoplastic changes in ulcerative colitis. Studies have confirmed that AI-assisted colonoscopy significantly increases adenoma detection rates, thereby reducing the incidence of interval colorectal cancer. Furthermore, AI-powered advanced endoscopy allows for a more objective assessment of mucosal and histological healing in IBD, helping predict outcomes and advancing precision medicine in its management. This narrative review comprehensively analyzes AI’s role in advanced endoscopic imaging, highlighting its impact on optical diagnosis in both upper and lower GI pathologies. It explores the integration of multimodal AI approaches, which combine imaging data with clinical and molecular biomarkers, to enhance diagnostic precision. Additionally, it discusses current challenges, including the need for multicenter validation, standardization of AI algorithms, and ethical considerations for clinical implementation. Future perspectives emphasize the necessity for high-quality prospective studies to validate AI’s real-world applicability and long-term benefits in endoscopic practice.