AMAlexandre Moreau-Gaudry

Alexandre Moreau-Gaudry

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Alexandre Moreau-Gaudry has a total of 33 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Explainable graph neural networks for early prediction of in-hospital mortality: a case study (Preprint)

    by Paul-Antoine Beaudoin, Thomas Boulier, Christophe Cancé, Svetlana Artemova, Sophie Achard, Jean-Luc Bosson, Alexandrre Moreau-Gaudry

    BACKGROUND Early prediction of in-hospital mortality is crucial for optimizing patient management and resource allocation. Traditional clinical risk scores remain widely used but are limited by their reliance on predefined variables and restricted applicability across clinical settings. Machine learning approaches have improved predictive performance by leveraging large-scale electronic health record data, while Graph Neural Networks (GNNs) offer additional advantages by modeling the complex relationships among patients, diagnoses, medications, and procedures. Despite these advances, the limited transparency of machine learning models remains a major obstacle to their adoption in clinical practice. In particular, local explainability, which aims to identify the factors driving individual predictions, is essential for supporting high-stakes medical decisions. Although numerous explainable artificial intelligence (XAI) methods have been proposed, robust and standardized approaches for evaluating the faithfulness, robustness, and clinical relevance of explanations remain scarce, hindering reliable comparison and deployment of explainable clinical AI systems. OBJECTIVE This study aims at extending a previously developed graph-based model for early in-hospital mortality prediction into a fully explainable framework. Specifically, we aimed to provide patient-level explanations for model predictions and evaluate several explanations methods through a quantitative framework. METHODS Building on our previous work, in which a heterogeneous graph was derived from a real-world clinical data warehouse and a GNN was trained to predict in-hospital mortality using only data available during the first 2 days of hospitalization, we developed an explainability framework completing the prediction pipeline. We introduced MCTS (Monte Carlo Tree Search)-Explainer and GNNExplainer-c, which extends the GNNexplainer method, two local explanation methods designed to identify the patient-specific clinical factors and graph structures driving predictions in our use case. Explanations were quantitatively evaluated using fidelity-based metrics, compared to state-of-the-art graph explainability methods, and further submitted to a clinical expert to evaluate end-user acceptability. RESULTS MCTS-Explainer achieved the highest fidelity-based score (characterization score = 0.53, 95% CI 0.52-0.54) while allowing more concise explanatory subgraphs and a shorter computation time. Qualitative review by an expert clinician confirmed acceptability of the results due to their concise and visual nature. CONCLUSIONS By extending a validated graph-based mortality prediction model with explanation generation and evaluation components, this study provides a comprehensive end-to-end framework for explainable clinical AI. MCTS-Explainer outperformed benchmark methods on quantitative evaluation metrics and was submitted to a medical expert for first qualitative assessments. These results demonstrate that graph-based models can support not only accurate prediction but also the feasibility of transparent and assessable decision support, promoting the trustworthy adoption of AI in healthcare.

  • Open Access

    Development of Indirect Health Data Linkage on Health Product Use and Care Trajectories in France: Systematic Review

    by Florence Ranchon, Sébastien Chanoine, Sophie Lambert-Lacroix, Jean-Luc Bosson, Alexandre Moreau-Gaudry, Pierrick Bedouch

    Background: European national disparities in the integration of data linkage (ie, being able to match patient data between databases) into routine public health activities were recently highlighted. In France, the claims database covers almost the whole population from birth to death, offering a great research potential for data linkage. As the use of a common unique identifier to directly link personal data is often limited, linkage with a set of indirect key identifiers has been developed, which is associated with the linkage quality challenge to minimize errors in linked data. Objective: The aim of this systematic review is to analyze the type and quality of research publications on indirect data linkage on health product use and care trajectories in France. Methods: A comprehensive search for all papers published in PubMed/Medline and Embase databases up to December 31, 2022, involving linked French database focusing on health products use or care trajectories was realized. Only studies based on the use of indirect identifiers were included (ie, without a unique personal identifier available to easily link the databases). A descriptive analysis of data linkage with quality indicators and adherence to the Bohensky framework for evaluating data linkage studies was also realized. Results: In total, 16 papers were selected. Data linkage was performed at the national level in 7 (43.8%) cases or at the local level in 9 (56.2%) studies. The number of patients included in the different databases and resulting from data linkage varied greatly, respectively, from 713 to 75,000 patients and from 210 to 31,000 linked patients. The diseases studied were mainly chronic diseases and infections. The objectives of the data linkage were multiple: to estimate the risk of adverse drug reactions (ADRs; n=6, 37.5%), to reconstruct the patient’s care trajectory (n=5, 31.3%), to describe therapeutic uses (n=2, 12.5%), to evaluate the benefits of treatments (n=2, 12.5%), and to evaluate treatment adherence (n=1, 6.3%). Registries are the most frequently linked databases with French claims data. No studies have looked at linking with a hospital data warehouse, a clinical trial database, or patient self-reported databases. The linkage approach was deterministic in 7 (43.8%) studies, probabilistic in 4 (25.0%) studies, and not specified in 5 (31.3%) studies. The linkage rate was mainly from 80% to 90% (reported in 11/15, 73.3%, studies). Adherence to the Bohensky framework for evaluating data linkage studies showed that the description of the source databases for the linkage was always performed but that the completion rate and accuracy of the variables to be linked were not systematically described. Conclusions: This review highlights the growing interest in health data linkage in France. Nevertheless, regulatory, technical, and human constraints remain major obstacles to their deployment. The volume, variety, and validity of the data represent a real challenge, and advanced expertise and skills in statistical analysis and artificial intelligence are required to treat these big data.


  • Open Access

    Development of Indirect Health Data Linkage on Health Product Use and Care Trajectories in France: Systematic Review (Preprint)

    by Florence Ranchon, Sébastien Chanoine, Sophie Lambert-Lacroix, Jean-Luc Bosson, Alexandre Moreau-Gaudry, Pierrick Bedouch

    BACKGROUND European national disparities in the integration of data linkage (ie, being able to match patient data between databases) into routine public health activities were recently highlighted. In France, the claims database covers almost the whole population from birth to death, offering a great research potential for data linkage. As the use of a common unique identifier to directly link personal data is often limited, linkage with a set of indirect key identifiers has been developed, which is associated with the linkage quality challenge to minimize errors in linked data. OBJECTIVE The aim of this systematic review is to analyze the type and quality of research publications on indirect data linkage on health product use and care trajectories in France. METHODS A comprehensive search for all papers published in PubMed/Medline and Embase databases up to December 31, 2022, involving linked French database focusing on health products use or care trajectories was realized. Only studies based on the use of indirect identifiers were included (ie, without a unique personal identifier available to easily link the databases). A descriptive analysis of data linkage with quality indicators and adherence to the Bohensky framework for evaluating data linkage studies was also realized. RESULTS In total, 16 papers were selected. Data linkage was performed at the national level in 7 (43.8%) cases or at the local level in 9 (56.2%) studies. The number of patients included in the different databases and resulting from data linkage varied greatly, respectively, from 713 to 75,000 patients and from 210 to 31,000 linked patients. The diseases studied were mainly chronic diseases and infections. The objectives of the data linkage were multiple: to estimate the risk of adverse drug reactions (ADRs; n=6, 37.5%), to reconstruct the patient’s care trajectory (n=5, 31.3%), to describe therapeutic uses (n=2, 12.5%), to evaluate the benefits of treatments (n=2, 12.5%), and to evaluate treatment adherence (n=1, 6.3%). Registries are the most frequently linked databases with French claims data. No studies have looked at linking with a hospital data warehouse, a clinical trial database, or patient self-reported databases. The linkage approach was deterministic in 7 (43.8%) studies, probabilistic in 4 (25.0%) studies, and not specified in 5 (31.3%) studies. The linkage rate was mainly from 80% to 90% (reported in 11/15, 73.3%, studies). Adherence to the Bohensky framework for evaluating data linkage studies showed that the description of the source databases for the linkage was always performed but that the completion rate and accuracy of the variables to be linked were not systematically described. CONCLUSIONS This review highlights the growing interest in health data linkage in France. Nevertheless, regulatory, technical, and human constraints remain major obstacles to their deployment. The volume, variety, and validity of the data represent a real challenge, and advanced expertise and skills in statistical analysis and artificial intelligence are required to treat these big data.

  • Open Access

    Evaluation of the clinical benefit of an electromagnetic navigation system for CT-guided interventional radiology procedures in the thoraco-abdominal region compared with conventional CT guidance (CTNAV II): study protocol for a randomised controlled trial

    by Rc. Rouchy, A. Moreau-Gaudry, E. Chipon, S. Aubry, L. Pazart, B. Lapuyade, M. Durand, M. Hajjam, S. Pottier, B. Renard, et al.

    Interventional radiology includes a range of minimally invasive image-guided diagnostic and therapeutic procedures that have become routine clinical practice. Each procedure involves a percutaneous needle insertion, often guided using computed tomography (CT) because of its availability and usability. However, procedures remain complicated, in particular when an obstacle must be avoided, meaning that an oblique trajectory is required. Navigation systems track the operator’s instruments, meaning the position and progression of the instruments are visualised in real time on the patient’s images. A novel electromagnetic navigation system for CT-guided interventional procedures (IMACTIS-CT®) has been developed, and a previous clinical trial demonstrated improved needle placement accuracy in navigation-assisted procedures. In the present trial, we are evaluating the clinical benefit of the navigation system during the needle insertion step of CT-guided procedures in the thoraco-abdominal region. This study is designed as an open, multicentre, prospective, randomised, controlled interventional clinical trial and is structured as a standard two-arm, parallel-design, individually randomised trial. A maximum of 500 patients will be enrolled. In the experimental arm (navigation system), the procedures are carried out using navigation assistance, and in the active comparator arm (CT), the procedures are carried out with conventional CT guidance. The randomisation is stratified by centre and by the expected difficulty of the procedure. The primary outcome of the trial is a combined criterion to assess the safety (number of serious adverse events), efficacy (number of targets reached) and performance (number of control scans acquired) of navigation-assisted, CT-guided procedures as evaluated by a blinded radiologist and confirmed by an expert committee in case of discordance. The secondary outcomes are (1) the duration of the procedure, (2) the satisfaction of the operator and (3) the irradiation dose delivered, with (4) subgroup analysis according to the expected difficulty of the procedure, as well as an evaluation of (5) the usability of the device. This trial addresses the lack of published high-level evidence studies in which navigation-assisted CT-guided interventional procedures are evaluated. This trial is important because it addresses the problems associated with conventional CT guidance and is particularly relevant because the number of interventional radiology procedures carried out in routine clinical practice is increasing. ClinicalTrials.gov identifier: NCT01896219 . Registered on 5 July 2013.