Artificial intelligence systems for endoscopic diagnosis of autoimmune atrophic gastritis: opportunities and limits

Chen et al. have highlighted several practical implications of their novel artificial intelligence (AI) system for diagnosing autoimmune atrophic gastritis (AIG) [1]. This work raises questions regarding the applicability of this AI model, which is based exclusively on gastric atrophy and the presence of Helicobacter pylori activity. Indeed, the authors focused their attention on gastric atrophy. However, the histopathological agreement on gastric atrophy is known to be lower than that for intestinal metaplasia [2], and the detection of intestinal metaplasia using electronic chromoendoscopy systems achieves high accuracy rates, approximately 85% [3]. Another important, yet overlooked, aspect is that the presence of atrophy without intestinal metaplasia has a lower risk of developing into gastric cancer. Similarly, the presence of gastric pseudopyloric metaplasia alone, which is a hallmark for an AIG diagnosis, is not associated with gastric cancer risk [4]. Thus, focusing only on gastric atrophy, could impact patient management.

Another point is the definition of AIG etiology. Currently, the greater challenge lies in distinguishing between AIG and H. pylori-related atrophic gastritis, as the main differentiating criteria are the presence or absence of anti-parietal cell antibodies (APCA) and antral involvement. According to the definition, AIG is “corpus-restricted,” with atrophy limited to the corpus, sparing the antrum. AI models based on gastric site evaluation can provide substantial diagnostic support. However, as demonstrated in a recent study, the eradication of H. pylori infection may lead to antral mucosal healing, which could mimic AIG [5]. To improve the performance of these systems for future real-time development, it would be beneficial to incorporate not only endoscopic images but also laboratory data (such as APCA, anti-intrinsic factor antibodies, and H. pylori immunoglobulin G antibodies). The integration of endoscopic, clinical/laboratory data, and electronic chromoendoscopy could significantly support an early etiological diagnosis of atrophic gastritis, even in real-time AI applications.

Publication History

Article published online:
22 April 2025

• References

• 1 Chen S, Xu L, Yan L. et al. A novel endoscopic artificial intelligence system to assist in the diagnosis of autoimmune gastritis: a multicenter study. Endoscopy 2024;
• 2 Leja M, Funka K, Janciauskas D. et al. Interobserver variation in assessment of gastric premalignant lesions: higher agreement for intestinal metaplasia than for atrophy. Eur J Gastroenterol Hepatol 2013; 25: 694-699
• 3 Pimentel-Nunes P, Libânio D, Lage J. et al. A multicenter prospective study of the real-time use of narrow-band imaging in the diagnosis of premalignant gastric conditions and lesions. Endoscopy 2016; 48: 723-730
• 4 Dilaghi E, Baldaro F, Pilozzi E. et al. Pseudopyloric metaplasia is not associated with the development of gastric cancer. Am J Gastroenterol 2021; 116: 1859-1867
• 5 Dottori L, Palumbo C, Dilaghi E. et al. Antral mucosa healing at long-term follow-up in patients with corpus atrophic gastritis and concomitant antral gastritis may mimic autoimmune gastritis. Dig Liver Dis 2024;