Santé, signal et apprentissage automatique
en collaboration avec S. Allassonnière, J. Stirnemann, Y. Ville, A. Pierangelo, G. Jauvion, E. Blondiaux, H. Amadou-Boubacar, G. Solelhac, C. Marini, I. Ben M’Barek (thèse 2023–), I. Kamal (thèse 2022–), A. Oulad Ali (thèse 2026–), R. Besson (thèse 2016–2019), F. Logé-Munerel (thèse 2017–2021), F. Gaudfernau (thèse 2020–2023), M. Brigham (postdoc 2017–2019) et T. Galtier (postdoc 2019–2020)
La prise en charge périnatale repose sur des signaux et des images (rythme cardiaque fœtal, échographie, IRM, imagerie optique) dont l’interprétation est difficile, souvent urgente et variable d’un clinicien à l’autre. Avec des obstétriciens, nous développons des outils d’apprentissage automatique pour aider ces décisions et nous les validons sur des données cliniques.
Surveillance fœtale pendant l’accouchement
Avec I. Ben M’Barek (thèse 2023–), J. Stirnemann et G. Jauvion, nous analysons la cardiotocographie (CTG) enregistrée pendant le travail pour détecter en temps réel l’hypoxie fœtale. Nous avons développé DeepCTG® 2.0, un modèle d’apprentissage profond qui détecte l’acidémie néonatale à partir de la CTG (Computers in Biology and Medicine 2025), étudié l’intégration de variables cliniques dans le système (European Journal of Obstetrics & Gynecology and Reproductive Biology 2025) et évalué, dans une étude randomisée, la collaboration entre cliniciens et algorithme pour l’interprétation de la CTG (npj Digital Medicine 2026). Une revue en français fait l’état des lieux de l’analyse automatisée de la CTG, de ses controverses et de ses perspectives (Gynécologie Obstétrique Fertilité & Sénologie 2026).
Imagerie polarimétrique du col de l’utérus
Avec A. Pierangelo, I. Kamal (thèse 2022–) et A. Oulad Ali (thèse 2026–), nous développons des outils d’apprentissage profond pour l’imagerie polarimétrique de Mueller, une modalité d’imagerie optique sensible à la microstructure des tissus. Nos travaux actuels segmentent les tissus du col dans des images de colposcopie polarimétrique de Mueller, une première étape vers des biomarqueurs optiques du risque d’accouchement prématuré (Proc. SPIE 2026).
Diagnostic prénatal et imagerie fœtale
Avec R. Besson (thèse 2016–2019), S. Allassonnière et les obstétriciens Y. Ville et J. Stirnemann (hôpital Necker), nous avons conçu un outil de diagnostic séquentiel des maladies fœtales rares qui combine un modèle expert et des données, et optimise l’enchaînement des examens par apprentissage par renforcement fondé sur un modèle (ICAART 2020). Sa validation clinique comme assistant de diagnostic en temps réel pour l’échographie fœtale (Ultrasound in Obstetrics & Gynecology 2021–2023) et un brevet (2024) ont accompagné son transfert vers Sonio, start-up issue de ces travaux et rachetée par Samsung Medison en 2024. Avec F. Gaudfernau (thèse 2020–2023), S. Allassonnière et E. Blondiaux, nous avons développé des modèles de déformations difféomorphes multiéchelles qui produisent des transformations d’images réalistes, appliqués à la modélisation de la croissance du cerveau fœtal (SPIE Medical Imaging 2023, Journal of Mathematical Imaging and Vision 2025).
Projets antérieurs : diabète, sommeil et données de santé (2017–2021)
Avec F. Logé-Munerel (thèse 2017–2021, CIFRE Air Liquide) et H. Amadou-Boubacar, nous avons utilisé la programmation dynamique approchée pour construire des questionnaires de santé adaptatifs (i-com 2021) et l’apprentissage par renforcement pour améliorer le conseil de bolus d’insuline des patients atteints de diabète de type 1 (KDD Workshop on Applied Data Science for Healthcare 2020). Avec M. Brigham (postdoc 2017–2019) et C. Marini, nous avons associé une blockchain à l’apprentissage automatique dans une étude sur le sommeil et, avec G. Solelhac, représenté des enregistrements de polysomnographie comme des trajectoires de faible dimension (Sleep 2019). Avec T. Galtier (postdoc 2019–2020), nous avons travaillé sur la valorisation des données et l’apprentissage fédéré pour les données de santé dans le projet HealthChain.
Les méthodes d’apprentissage par renforcement sur lesquelles reposent l’outil de diagnostic et les questionnaires sont présentées sur la page apprentissage par renforcement.
Publications
2026 (4)
-
Analyse automatisée de la cardiotocographie en intrapartum : état des lieux, controverses et perspectives. Gynécologie Obstétrique Fertilité & Sénologie, 54(6):342–349, 2026.
PDFDOI
BibTeX
@article{benmbarek26a, author = {Ben M'Barek, I. and Holmström, E. and Ceccaldi, P.-F. and Michel, J. and Vitrou, J. and Le Pennec, E. and Stirnemann, J.}, title = {Analyse automatisée de la cardiotocographie en intrapartum : état des lieux, controverses et perspectives}, journal = {Gynécologie Obstétrique Fertilité \& Sénologie}, number = {6}, volume = {54}, pages = {342--349}, year = {2026}, doi = {10.1016/j.gofs.2026.01.006} } -
Randomised study of human machine collaboration for cardiotocography interpretation during labour. npj Digital Medicine, 9(1):365, 2026.
PDFDOIHAL
BibTeX
@article{benmbarek26b, author = {Ben M'Barek, I. and Ben M'Barek, B. and Jauvion, G. and Whelehan, V. and Papageorghiou, A. and Le Pennec, E. and Stirnemann, J.}, title = {Randomised study of human machine collaboration for cardiotocography interpretation during labour}, journal = {npj Digital Medicine}, number = {1}, volume = {9}, pages = {365}, year = {2026}, doi = {10.1038/s41746-026-02556-y} }
-
Deep learning-based semantic segmentation of cervical tissue in Mueller polarimetric colposcopy images for preterm risk assessment. Polarized Light and Optical Angular Momentum for Biomedical Diagnostics 2026, Proc. SPIE, 2026.
DOI
BibTeX
@inproceedings{kamal26a, author = {Kamal, I. and Park, J. and Vanel, J.-C. and Luton, D. and Debras, E. and Le Pennec, E. and Pierangelo, A.}, title = {Deep learning-based semantic segmentation of cervical tissue in Mueller polarimetric colposcopy images for preterm risk assessment}, booktitle = {Polarized Light and Optical Angular Momentum for Biomedical Diagnostics 2026}, series = {Proc. SPIE}, year = {2026}, doi = {10.1117/12.3079701} } -
Towards optical biomarkers of preterm birth: deep learning segmentation of cervical tissue using Mueller matrix polarimetry. Unconventional Optical Imaging V, Proc. SPIE, 2026.
DOI
BibTeX
@inproceedings{kamal26b, author = {Kamal, I. and Park, J. and Vanel, J.-C. and Luton, D. and Debras, E. and Gandji, A. and Le Pennec, E. and Pierangelo, A.}, title = {Towards optical biomarkers of preterm birth: deep learning segmentation of cervical tissue using Mueller matrix polarimetry}, booktitle = {Unconventional Optical Imaging V}, series = {Proc. SPIE}, year = {2026}, doi = {10.1117/12.3100057} }
2025 (5)
-
Wavelet-Based Multiscale Flow For Realistic Image Deformation in the Large Diffeomorphic Deformation Model Framework. Journal of Mathematical Imaging and Vision, 67(2):10, 2025.
PDFPrépublicationDOIarXivHAL
BibTeX
@article{gaudfernau24, author = {Gaudfernau, F. and Blondiaux, E. and Allassonnière, S. and Le Pennec, E.}, title = {Wavelet-Based Multiscale Flow For Realistic Image Deformation in the Large Diffeomorphic Deformation Model Framework}, journal = {Journal of Mathematical Imaging and Vision}, number = {2}, volume = {67}, pages = {10}, year = {2025}, doi = {10.1007/s10851-024-01219-5}, eprint = {2501.18211}, archiveprefix = {arXiv} } -
Integration of clinical features in a computerized cardiotocography system to predict severe newborn acidemia. European Journal of Obstetrics & Gynecology and Reproductive Biology, 307:78-83, 2025.
PDFDOI
BibTeX
@article{benmarek25, author = {Menzhulina, E. and Vitrou, J. and Merrer, J. and Holmström, E. and Ait Amara, I. and Le Pennec, E. and Stirnemann, J. and Ben M'Barek, I.}, title = {Integration of clinical features in a computerized cardiotocography system to predict severe newborn acidemia}, journal = {European Journal of Obstetrics \& Gynecology and Reproductive Biology}, volume = {307}, pages = {78-83}, year = {2025}, doi = {10.1016/j.ejogrb.2025.01.030} } -
DeepCTG® 2.0: Development and validation of a deep learning model to detect neonatal acidemia from cardiotocography during labor. Computers in Biology and Medicine, 184:109448, 2025.
PDFDOI
BibTeX
@article{benmarek24, author = {Ben M'Barek, I. and Jauvion, G. and Merrer, J. and Koskas, M. and Sibony, O. and Ceccaldi, P.-F. and Le Pennec, E. and Stirnemann, J.}, title = {DeepCTG® 2.0: Development and validation of a deep learning model to detect neonatal acidemia from cardiotocography during labor}, journal = {Computers in Biology and Medicine}, volume = {184}, pages = {109448}, year = {2025}, doi = {10.1016/j.compbiomed.2024.109448} }
-
Human-machine collaboration in intrapartum care: A randomized study of computerized assistance for clinical cardiotocography interpretation and prediction of perinatal acidemia. Gynécologie Obstétrique Fertilité & Sénologie, 53(11):605-606, 2025.
PDFDOI
BibTeX
@inproceedings{benmarek25c, author = {I. {Ben M'Barek} and B. {Ben M'Barek} and G. Jauvion and V. Whelehan and A. Papageorghiou and E. {Le Pennec} and J. Stirnemann}, title = {Human-machine collaboration in intrapartum care: A randomized study of computerized assistance for clinical cardiotocography interpretation and prediction of perinatal acidemia}, booktitle = {Gynécologie Obstétrique Fertilité \& Sénologie}, number = {11}, volume = {53}, pages = {605-606}, year = {2025}, doi = {10.1016/j.gofs.2025.09.030} } -
DeepCTG® 2.0: development and validation of a deep learning model to detect fetal acidemia from cardiotocography during labor. European Journal of Obstetrics & Gynecology and Reproductive Biology, 313(S1):114431, 2025.
PDFDOI
BibTeX
@inproceedings{benmarek25b, author = {Ben M'Barek, I. and Jauvion, G. and Stirnemann, J. and Le Pennec, E.}, title = {DeepCTG® 2.0: development and validation of a deep learning model to detect fetal acidemia from cardiotocography during labor}, booktitle = {European Journal of Obstetrics \& Gynecology and Reproductive Biology}, number = {S1}, volume = {313}, pages = {114431}, year = {2025}, doi = {10.1016/j.ejogrb.2025.114431} }
2024 (1)
-
Real-time diagnosis aid method and decision-support for medical diagnosis to a user of a medical system. US11881309B2, 2024.
PDF
BibTeX
@patent{allassonniere21, author = {Allassonnière, S. and Besson, R. and Le Pennec, E.}, title = {Real-time diagnosis aid method and decision-support for medical diagnosis to a user of a medical system}, number = {US11881309B2}, year = {2024} }
2023 (3)
-
Development and clinical validation of real-time artificial intelligence diagnostic companion for fetal ultrasound examination. Ultrasound in Obstetrics & Gynecology, 62(3):353-360, 2023.
PDFDOI
BibTeX
@article{stirnemann23, author = {Stirnemann, J. and Besson, R. and Spaggiari, E. and Rojo, S. and Logé, F. and Peyro-Saint-Paul, H. and Allassonnière, S. and Le Pennec, E. and Hutchinson, C. and Sebire, N.J. and Ville, Y.}, title = {Development and clinical validation of real-time artificial intelligence diagnostic companion for fetal ultrasound examination}, journal = {Ultrasound in Obstetrics \& Gynecology}, number = {3}, volume = {62}, pages = {353-360}, year = {2023}, doi = {10.1002/uog.26242} }
-
A multiscale algorithm for computing realistic image transformations in the LDDMM framework - Application to the modelling of fetal brain growth. SPIE Medical Imaging 2023: Image Processing, Proc. SPIE, 12464:1246404, 2023.
PDF
BibTeX
@inproceedings{gaudfernau23, author = {Gaudfernau, F. and Allassonnière, S. and Le Pennec, E.}, title = {A multiscale algorithm for computing realistic image transformations in the LDDMM framework - Application to the modelling of fetal brain growth}, booktitle = {SPIE Medical Imaging 2023: Image Processing}, series = {Proc. SPIE}, volume = {12464}, pages = {1246404}, year = {2023} }
-
Un algorithme multiéchelle pour déformer les objets de façon réaliste - application à la modélisation de la croissance du cerveau foetal. GRETSI, 2023.
PDF
BibTeX
@inproceedings{gaudfernau23b, author = {Gaudfernau, F. and Allassonnière, S. and Le Pennec, E.}, title = {Un algorithme multiéchelle pour déformer les objets de façon réaliste - application à la modélisation de la croissance du cerveau foetal}, booktitle = {GRETSI}, year = {2023} }
2022 (1)
-
Clinical validation of a real‐time AI diagnostic companion for fetal ultrasound examination. Ultrasound in Obstetrics & Gynecology, 60(S1):14-15, 2022.
PDF
BibTeX
@inproceedings{stirnemann22, author = {Stirnemann, J. and Besson, R. and Spaggiari, E. and Rojo, S. and Logé, F. and Peyro-Saint-Paul, H. and Allassonnière, S. and Le Pennec, E. and Hutchinson, C. and Sebire, N.J. and Ville, Y.}, title = {Clinical validation of a real‐time AI diagnostic companion for fetal ultrasound examination}, booktitle = {Ultrasound in Obstetrics \& Gynecology}, number = {S1}, volume = {60}, pages = {14-15}, year = {2022} }
2021 (2)
-
Description and clinical validation of a real-time AI diagnostic companion for fetal ultrasound examination. Ultrasound in Obstetrics & Gynecology, 58(S1):169-170, 2021.
PDFPrépublicationDOI
BibTeX
@inproceedings{stirnemann21, author = {Stirnemann, J. and Besson, R. and Spaggiari, E. and Bourgon, N. and Rojo, S. and Logé, F. and Peyro-Saint-Paul, H. and Allassonnière, S. and Le Pennec, E. and Ville, Y.}, title = {Description and clinical validation of a real-time AI diagnostic companion for fetal ultrasound examination}, booktitle = {Ultrasound in Obstetrics \& Gynecology}, number = {S1}, volume = {58}, pages = {169-170}, year = {2021}, doi = {10.1002/uog.24291} } -
A Model-Based Reinforcement Learning Approach for a Rare Disease Diagnostic Task. Proceedings of Entropy 2021: The Scientific Tool of the 21st Century, 2021.
DOI
BibTeX
@inproceedings{Besson21, author = {Besson, R. and Le Pennec, E. and Allassonnière, S.}, title = {A Model-Based Reinforcement Learning Approach for a Rare Disease Diagnostic Task}, booktitle = {Proceedings of Entropy 2021: The Scientific Tool of the 21st Century}, year = {2021}, doi = {10.3390/entropy2021-09749} }
2020 (4)
-
Intelligent Questionnaires Using Approximate Dynamic Programming. i-com, 19(3):227-237, 2020.
PDFDOI
BibTeX
@article{Loge21, author = {Logé, F. and Le Pennec, E. and Amadou-Boubacar, H.}, title = {Intelligent Questionnaires Using Approximate Dynamic Programming}, journal = {i-com}, number = {3}, volume = {19}, pages = {227-237}, year = {2020}, doi = {10.1515/icom-2020-0022} }
-
Challenging common bolus advisor for self-monitoring type-I diabetes patients using Reinforcement Learning. 2020 KDD Workshop on Applied Data Science for Healthcare, 2020.
PDFarXiv
BibTeX
@inproceedings{LogeDSHealh20, author = {Logé, F. and Le Pennec, E. and Amadou-Boubacar, H.}, title = {Challenging common bolus advisor for self-monitoring type-I diabetes patients using Reinforcement Learning}, booktitle = {2020 KDD Workshop on Applied Data Science for Healthcare}, year = {2020}, eprint = {2007.11880}, archiveprefix = {arXiv} } -
Adaptive predictive-questionnaire by approximate dynamic programming. Workshop on User-Centered Artificial Intelligence (UCAI), Mensch und Computer 2020, 2020.
PDFDOI
BibTeX
@inproceedings{LogeUCAI20, author = {Logé, F. and Le Pennec, E. and Amadou-Boubacar, H.}, title = {Adaptive predictive-questionnaire by approximate dynamic programming}, booktitle = {Workshop on User-Centered Artificial Intelligence (UCAI), Mensch und Computer 2020}, year = {2020}, doi = {10.18420/muc2020-ws111-264} } -
Optimization of a Sequential Decision Making Problem for a Rare Disease Diagnostic Application. ICAART 2020, pp. 475–482, 2020.
PDFDOI
BibTeX
@inproceedings{BessonIcaart20, author = {Besson, R. and Le Pennec, E. and Spaggiari, E. and Neuraz, A. and Stirnemann, J. and Allassonnière, S.}, title = {Optimization of a Sequential Decision Making Problem for a Rare Disease Diagnostic Application}, booktitle = {ICAART 2020}, pages = {475--482}, year = {2020}, doi = {10.5220/0008938804750482} }
2019 (3)
-
Learning from both experts and data. Entropy, 21(12):1208, 2019.
PDFDOIarXivHAL
BibTeX
@article{Besson19, author = {Besson, R. and Le Pennec, E. and Allassonnière, S.}, title = {Learning from both experts and data}, journal = {Entropy}, number = {12}, volume = {21}, pages = {1208}, year = {2019}, doi = {10.3390/e21121208}, eprint = {1910.09043}, archiveprefix = {arXiv} }
-
Representation Of Polysomnography Recordings As Low Dimensional Trajectories. Sleep, 42:A128, 33rd Annual Meeting of the Associated Professional Sleep Societies, 2019.
PDFDOI
BibTeX
@inproceedings{Solelhac19, author = {Solelhac, G. and Brigham, M. and Bouchequet, P. and Andrillon, Th. and Chennaoui, M. and Le Pennec, E. and Rey, M. and Léger, D.}, title = {Representation Of Polysomnography Recordings As Low Dimensional Trajectories}, booktitle = {Sleep}, volume = {42}, pages = {A128}, note = {33rd Annual Meeting of the Associated Professional Sleep Societies}, year = {2019}, doi = {10.1093/sleep/zsz067.314} }
-
Représentation d'enregistrements de polysomnographie en trajectoires de basses dimensions dans l'espace latent. Médecine du Sommeil, 16(1):56, 2019.
PDFDOI
BibTeX
@inproceedings{Solelhac19b, author = {Solelhac, G. and Brigham, M. and Bouchequet, P. and Andrillon, Th. and Chennaoui, M. and Le Pennec, E. and Rey, M. and Léger, D.}, title = {Représentation d'enregistrements de polysomnographie en trajectoires de basses dimensions dans l'espace latent}, booktitle = {Médecine du Sommeil}, number = {1}, volume = {16}, pages = {56}, year = {2019}, doi = {10.1016/j.msom.2019.01.106} }
2018 (2)
-
Une nouvelle représentation de la polysomnographie par une technique de machine learning nonsupervisée. Médecine du Sommeil, 15(1):49, Résumés du Congrès du Sommeil. Marseille, 23-25 novembre 2017, 2018.
PDFDOI
BibTeX
@inproceedings{Solelhac18nouvelle, author = {Solelhac, G. and Brigham, M. and Marini, C. and Bouchequet, P. and Chennaoui, M. and Le Pennec, E. and Léger, D.}, title = {Une nouvelle représentation de la polysomnographie par une technique de machine learning nonsupervisée}, booktitle = {Médecine du Sommeil}, number = {1}, volume = {15}, pages = {49}, note = {Résumés du Congrès du Sommeil. Marseille, 23-25 novembre 2017}, year = {2018}, doi = {10.1016/j.msom.2018.01.132} }
-
A Model-Based Reinforcement Learning Approach for a Rare Disease Diagnostic Task. Preprint, 2018.
PDFarXiv
BibTeX
@article{Besson18, author = {Besson, R. and Le Pennec, E. and Allassonnière, S. and Stirnemann, J. and Spaggiari, E. and Neuraz, A.}, title = {A Model-Based Reinforcement Learning Approach for a Rare Disease Diagnostic Task}, journaltitle = {Preprint}, year = {2018}, eprint = {1811.10112}, archiveprefix = {arXiv} }
2017 (1)
-
Optimisation d'arbre de décision pour un problème de détection précoce d'anomalies foetales. Journée de la SFdS, Avignon, 2017.
PDF
BibTeX
@inproceedings{BessonSfdS17, author = {Besson, R. and Le Pennec, E. and Allassonnière, S.}, title = {Optimisation d'arbre de décision pour un problème de détection précoce d'anomalies foetales}, booktitle = {Journée de la SFdS}, address = {Avignon}, year = {2017} }
Exposés
2026 (1)
- Reinforcement Learning: An Introduction and Some Results. Séminaire Probabilités et Statistiques, LMO, Laboratoire de Mathématiques d'Orsay, Université Paris-Saclay, Orsay, France, avril 2026. Transparents
2024 (1)
- Reinforcement Learning: An Introduction and Some Results. 9e Journée Statistique et Informatique pour la Science des Données à Paris-Saclay, IHES, Bures-sur-Yvette, France, avril 2024. TransparentsVidéo
2022 (1)
- Optimization of Sequential-Decision Problems in Health using Reinforcement Learning. MVA, ENS Paris-Saclay, ENS Paris-Saclay, Gif-sur-Yvette, France, septembre 2022. Transparents
2021 (1)
- Optimization of a Sequential Decision Making Problem for a Rare Disease Diagnostic Application. Seminaire Telecom, Telecom, Palaiseau, France, novembre 2021. Transparents
2019 (2)
- Optimization of a Sequential Decision Making Problem for a Rare Disease Diagnostic Application. ANR COSMOS, Inria Paris, Paris, France, novembre 2019. Transparents
- Optimization of a Sequential Decision Making Problem for a Rare Disease Diagnostic Application. Agro Paris, Agro Paris, Paris, France, octobre 2019. Transparents
2018 (1)
- Optimization of a Sequential Decision Making Problem for a Rare Disease Diagnostic Application. TSE, TSE, Toulouse, France, octobre 2018. Transparents