The Helicobacter pylori AI-clinician harnesses artificial intelligence to personalise H. pylori treatment recommendations

• Type: Article in International Scientific Journal
• Date: 2025

Authors

Nyssen, OP (Author) Other; Southern, J (Author) Other; Laponogov, I (Author) Other; Marco, AM (Author) Other; Cabeza-Segura, M (Author) Other; Martí, EJ (Author) Other; Castillo, J (Author) Other; Leja, M (Author) Other; Po¿aka, I (Author) Other; Carneiro, F (Author) Other; Figueiredo C (Author) FMUP; Ferreira, RM (Author) Other; Barros, R (Author) Other; Moreira, L (Author) Other; Cuatrecasas, M (Author) Other; Fernandez-Esparrach, G (Author) Other; Matysiak-Budnik, T (Author) Other; Martin, J (Author) Other; Jonaitis, L (Author) Other; Kup¿inskas, J (Author) Other; Jonaitis, P (Author) Other; Dinis-Ribeiro, M (Author) Other; Coimbra, M (Author) Other; Pereira, AC (Author) Other; Fontes, F (Author) Other; Spaander, MCW (Author) Other; Honing, J (Author) Other; Sedola, S (Author) Other; Pescino, JA (Author) Other; Maravic, Z (Author) Other; Martins, A (Author) Other; Veselkov, D (Author) Other; Gisbert, JP (Author) Other; Kanonnikoff, TF (Author) Other; Veselkov, K (Author) Other; Higgins, K (Author) Other

Journal

Title: Nature Communications

Vol. 10

ISSN: 2041-1723

Publisher: Springer Nature

Other information

Authenticus ID: P-019-MAP

DOI: 10.1038/s41467-025-61329-5

Abstract (EN): <jats:title>Abstract</jats:title> <jats:p> <jats:italic>Helicobacter pylori</jats:italic> (<jats:italic>H. pylori</jats:italic>) is the most common carcinogenic pathogen globally and the leading cause of gastric cancer. Here, we develop a reinforcement learning-based AI Clinician system to personalise treatment selection and evaluate its ability to improve eradication success compared to clinician-prescribed therapies. The model is trained and internally validated on 38,049 patients from the retrospective European Registry on <jats:italic>Helicobacter pylori</jats:italic> Management (Hp-EuReg), using independent state deep Q-learning (isDQN) to recommend optimal therapies based on patient characteristics such as age, sex, antibiotic allergies, country, and pre-treatment indication. In internal validation using real-world Hp-EuReg data, AI-recommended therapies achieve a 94.1% success rate (95% CI: 93.2¿95.0%) versus 88.1% (95% CI: 87.7¿88.4%) for clinician-prescribed therapies not aligned with AI suggestions¿an improvement of 6.0%. Results are replicated in an external validation cohort (<jats:italic>n</jats:italic>¿=¿7186), confirming generalisability. The AI system identifies optimal treatment strategies in key subgroups: 65% (<jats:italic>n</jats:italic>¿=¿24,923) are recommended bismuth-based therapies, and 15% (<jats:italic>n</jats:italic>¿=¿5898) non-bismuth quadruple therapies. Random forest modelling identifies region and concurrent medications as patient-specific drivers of AI recommendations. With nearly half the global population likely to contract <jats:italic>H. pylori</jats:italic>, this approach lays the foundation for future prospective clinical validation and shows the potential of AI to support clinical decision-making, enhance outcomes, and reduce gastric cancer burden.</jats:p>

Language: English

Type (Professor's evaluation): Scientific