Endoscopic examination plays a crucial role in the secondary prevention of gastric cancer. However, its effectiveness is often limited by the subjective nature of assessment and the risk of missing neoplastic lesions. The use of deep learning (DL) and machine learning (ML) models allows for a significant improvement in the quality of endoscopic diagnosis of early gastric cancer. Real-time systems increase the sensitivity of the examination and reduce skill-based differences between endoscopists. They also enable precise assessment of tumor invasion depth, which is essential for qualifying patients for minimally invasive treatment. Explainable artificial intelligence (XAI) methods and ML algorithms are gaining increasing importance as they enhance the transparency of the diagnostic process. Despite promising data, the full implementation of AI systems in endoscopy faces challenges such as susceptibility to imaging artifacts, the risk of model overfitting, and regulatory hurdles. A key issue remains the development of representative, multicenter training image databases and the conduct of prospective clinical trials, which may revolutionize standards of care. The aim of this article is to discuss the role of AI-assisted gastroscopy in the prevention of early gastric cancer, with particular emphasis on its potential in population-based screening programs. The paper presents current clinical data and analyzes the benefits and challenges associated with implementing these technologies into routine clinical practice.
Koronczok-Matusiak et al. (Mon,) studied this question.