Innovative research group dedicated to improving emergency triage and patient care through Artificial Intelligence. Led by Dr Edouard Lansiaux, we develop solutions that save lives.
Discover our projectsResearch projects
Patients enrolled
AI concordance
Partner centers
A progressive and complementary approach to revolutionize emergency triage
Single-center retrospective study (n=657, Lille University Hospital) demonstrating superiority of 3 AI models (NLP, LLM, JEPA) over standard nursing triage. Concordance of 90% vs 30% (p<0.001).
Publication: IEEE BDCAT 2025, JMIR Medical Informatics
OngoingExternal validation across 5 centers (CHU Lille, CH Douai, Denain, Maubeuge, Tenon AP-HP) with several thousand patients. Period: January-February 2026.
Type: Prospective observational study
Funding pendingProspective cluster-randomized bicentric trial (CHU Lille, CH Maubeuge) evaluating AI assistance under real-world conditions. Objective: concordance increase from 68% to 76%.
Funding requested: PHRC-I & doctoral grant
Completed - M2 dissertationRetraining of the 3 TIAEU architectures on 340,536 patients and dual-regime R1/R2 evaluation. Under clinical validity (R2), only URGENTIAPARSE clears the deployment threshold (κw 0.81). DES-MAS simulation + PSA economics (ROI 480%).
Approach: DES + MAS Simulation, 4D digital twins, CHEERS 2022
International collaborationRedirection application integrating AI to identify cases suitable for primary care. Estimated 15-20% reduction in emergency department overcrowding.
Partners: CHU Quebec, Mila, IVADO
Preliminary resultsProspective study comparing an LLM (Claude family) against call dispatchers and dispatch physicians at the SAS-Centre 15 on simulated scenarios (SimSamu corpus). First wave (30 scenarios): AI reaches physician level (κ 0.46 vs 0.40) with a cautious "safety-first" profile.
Center: SAS-Centre 15, Lille University Hospital
International collaborationInternational multicenter federated study evaluating an innovative triage effectiveness metric in emergency departments. Federated analysis with row-level data processed locally.
Partners: Lund Univ. 🇸🇪, Yale 🇺🇸, UVA 🇺🇸
PERCEPT'urg — What 141 French emergency professionals really think about their specialty.
This conference isn't about the AI you'd want in the ER. It's about the AI that's already there — or that's in clinical trials right now.
This communication presented at the Emergency 2026 Congress (SFMU, Paris) sets out the results of the TIAEU study: a retrospective concept study on 657 emergency visits evaluating the ability of an NLP + XGBoost model to predict the level of triage according to the FRENCH scale from the reason for consultation in natural language.
Lansiaux E, Leman M.
Lansiaux E, Jairi I, Zgaya-Biau H.
Lansiaux E.
Lansiaux E, Leman M, Ammi M.
Lansiaux E, Marx J, Tataru G.
Lansiaux E, Guerif Dubreucq E, Chan T.
Lansiaux E, Simonet A, Wiel E.
Lansiaux E, Azzouz R, Chazard E, Vromant A, Wiel E.
Lansiaux E, Azzouz R, Chazard E, Vromant A, Wiel E.
Lansiaux E, Arnaud E, Arrouy L, Auboiroux PH, Balaz PA, Baron MA, Depil-Duval A, Dubreucq-Guérif E, Dumontier T, Ellouze S, Gil-Jardine C, Gilbert A, Heidet M, Mpela AG, Lemaitre EL, Vromant A, Violeau M.
Lansiaux E.
Médecine de Catastrophe - Urgences Collectives, 9(2), 96-100
Lansiaux E, Baron MA, Vromant A.
At EUSEM 2025 in Vienna, this session explores how AI is transforming patient reception and triage in emergency departments.
Faced with increased ED activity, can AI one day supplement nursing expertise at the emergency department entrance?
Between technological innovation, operational pressure, and ethical challenges — the growing role of AI in emergency services in 2025.
No publications match this filter combination.
Triage accuracy improvement from 30% to 90%
Reduction of dangerous under-triage for patients
Optimization of hospital resources
International collaborations France-Canada-Sweden-USA