Research
My work at the CREATIS laboratory aims to improve patient care through effective interaction between clinicians and deep learning methods. It is organised into four areas.
Research areas
3D ultrasound imaging
Sparse array design, coded transmit sequences and image reconstruction. In the ANR SPARTECHUS project, carried out with the probe manufacturer IMASONIC, we showed that complementary coded sequences strongly increase the signal-to-noise ratio of sparse arrays without lowering the frame rate, in 2D and 3D, in particular for ultrafast Power Doppler imaging.
Cerebral emboli classification with transcranial Doppler
For stroke prevention, we developed interactive annotation and label propagation tools that let an expert interpret and improve the results of a deep neural network within minutes. This work, carried out with the company Atys Medical, also covers hybrid architectures (convolutional transformer) and model compression for frugal AI.
Personalised ventilation in intensive care
Analysing the chest CT of patients with acute respiratory distress syndrome (ARDS) makes it possible to measure local lung aeration and personalise mechanical ventilator settings. Our segmentation model, validated against experts, was made robust to scanners from the main manufacturers thanks to a method for generating synthetic patient images. This work is part of the Functional Imaging and Modeling of the Lungs group at CREATIS and of the IAVAI pre-maturation project.
Vascular malformations in MRI
Within the RHU COSY project on overgrowth syndromes, we segment vascular lesions in MRI using unsupervised learning and generative models, to give experts a pre-segmentation to correct and to follow lesion volume under treatment. A collaboration with Kitware integrated these models into 3D Slicer modules.
Funded projects
- ANR OPTIVENT, Optimizing Mechanical Ventilation through Personalized Digital Twin and Deep Learning (2026–2030). Partner of the “Digital twin generation on local clinical imaging data” work package; project led by Prof. Adrien Bartoli (University hospital of Clermont Ferrand, DRCI.)
- IAVAI, artificial intelligence for imaging-assisted ventilation (2024–2026). Pre-maturation, Impulse France 2030 programme (Inserm Transfert and Université Lyon 1). Co-leader with Prof. Maciej Orkisz.
- EvoluNet, Evolution of Neural Networks (2024–2025). Fédération Informatique de Lyon (CNRS). Co-leader with Prof. Guillaume Beslon (INSA Lyon, INRIA).
- SegLung COVID-CTPRED (2024). CREATIS internal call for projects. Leader.
- Arqus, Improving Mechanical Ventilation for ARDS (2022–2023). Arqus European University Alliance. Leader, in collaboration with Prof. Andreas Reske (Leipzig, Zwickau).
- ANR CARPUS (2022–2026). Ultrasound characterisation of carotid atherosclerotic plaque. Participant; project led by Pauline Muleki Seya (CNRS).
- ANR SPARTECHUS, Sparse array techniques for 3D medical ultrasound (2021–2025). Leader of the “optimisation and co-design” work package; project led by Prof. Hervé Liebgott (Lyon1 Universiy).
- RHU COSY, Cure Overgrowth Syndromes (2019–2025). Participant; project led by Prof. Guillaume Canaud (Necker Hospital).
- CAREMB (2019–2024). Auvergne-Rhône-Alpes Region. Participant; project led by Prof. Philippe Delachartre (INSA Lyon).
Technology transfer
- French patent application filed in December 2025 with the technology transfer office Pulsalys: method for generating synthetic computed tomography images.
- creaSDRA prototype software, for analysing CT scans of ARDS patients.
- Cooperation agreement between CREATIS and the Heinrich-Braun Hospital in Zwickau (Germany).
Supervision
Completed PhD theses
- Antoine Fraissenon (2020–2025), Phenotyping and radiological follow-up of PIK3CA patients. With Dr. Patrick Clarysse and Prof. Guillaume Canaud.
- Mohamed Tamraoui (2021–2024), Coded imaging sequences and reconstruction algorithm for ultrafast 3D ultrasound with sparse 2D arrays. With Prof. Hervé Liebgott.
- Yamil Vindas (2020–2023), Weakly-supervised learning for emboli characterization with Transcranial Doppler monitoring. With Prof. Philippe Delachartre and Kevin Guépié.
- Ludmilla Penarrubia (2019–2022), Quantification of pulmonary aeration in CT scans of patients with acute respiratory distress syndrome. With Prof. Jean-Christophe Richard and Prof. Maciej Orkisz.
Postdoctoral researcher and engineer
- Ludmilla Penarrubia, postdoctoral researcher (2024–2026): training strategies and data augmentation through hybrid image generation (IAVAI project).
- Alisa Kugusheva, research engineer (2024–2025): diffusion models for vascular lesion segmentation in MRI (RHU COSY).
Master’s internships and final-year projects
- Vincent Guardies (2025), Evolutionary Unconstraint Networks.
- Romane Milcent (2024), data augmentation for learning from pairs of ARDS CT images.
- Basile Caracalla (2024), Smooth volumetric copy-paste for ARDS CT database augmentation.
- Mario Ortiz Vargas (2024), final-year project with the Pontifical Xavierian University in Bogotá: lung segmentation tools for the COVID-CTPRED database.
- Maria Marquez (2023), modelling and semi-automatic segmentation of lungs in CT images.
- Fred Ngoga (2022), Self-supervised deep learning to improve thoracic CT-scan embedding.