← Back to projects
Ongoing - Hybrid simulation

EIMLIA-3M-TEU

AI Model Intelligence Evaluation - 3 Models for Emergency Triage

Project Description

EIMLIA-3M-TEU is a multidimensional comparative evaluation study of the three AI architectures developed in TIAEU. This project combines Process Mining, Discrete Event Simulation (DES), and Multi-Agent Systems to create a digital twin of the emergency department.

Approach

Hybrid DES + Multi-Agent Systems simulation

Process Mining

Automated patient pathway extraction

Models

TRIAGEMASTER (NLP)
URGENTIAPARSE (LLM)
EMERGINET (JEPA)

Indicators

Length of stay, mortality, resource allocation

Evaluated AI Models

NLP-TRIAGEMASTER

Doc2VecNeural Regression

Unsupervised text analysis with vector representations of verbatims.

LLM-URGENTIAPARSE

FlauBERTFine-tuningMultimodal Fusion

Pre-trained LLM with fusion of textual embeddings and vital signs.

JEPA-EMERGINET

Joint EmbeddingPredictive Architecture

JEPA architecture with specialized encoders and latent space learning.

Evaluation Dimensions

Technical Performance

Accuracy, sensitivity, specificity, AUC-ROC, F1-score

Organizational Impact

Length of stay, waiting time, room occupancy

Patient Safety

Under-triage rate, simulated mortality

Efficiency

Cost per patient, resource optimization

Ethics

MR-004 methodology with declaration N° 27797006 to the Health Data Hub.

Project Team

Dr Edouard Lansiaux

Principal Investigator
Lille University Hospital

Pr Emmanuel Chazard

Methodologist
METRICS ULR 2694

Pr Eric Wiel

Clinical Coordinator
METRICS ULR 2694 & Lille University Hospital

Pr Hayfa Zgaya-Biau

Process Mining Expert
METRICS ULR 2694 & CRIStAL UMR 9189

Pr Mehdi Ammi

Man-machine Interface Expert
LIASD

Dr Ramy Azzouz

AI in Healthcare Expertise
Lille University Hospital