Research Engineer
I build digital twins of the heart: models that reproduce a patient's heart from clinical examinations, and on which one can experiment freely. In the clinic as in research or industry, their usefulness comes down to two things, fidelity and speed. That is what I work on, somewhere between physics, machine learning and computing.
Estimating the parameters of an electromechanical model from a cine MRI and an ECG, with calibrated uncertainty: knowing what each examination reveals and what it leaves undetermined. The whole pipeline follows verification, validation and uncertainty quantification.
Simulating patients once, then rendering them as cine MRI, echocardiography, CT and ECG with generative models constrained by physics. Data that is consistent across modalities, with known ground truth and privacy-safe, to train and validate AI.
Describing the motion of the heart with a few regional transformations, learning the space of these motions and its geometry, then giving it physical dynamics. The same formalism summarises motion for inference and animates it for rendering.
3D electromechanical models of the heart, to understand cardiac function and help plan treatment for heart failure patients. Built with SOFA and run on Inria's national computing infrastructure.
A patient-specific preoperative planning tool for transcatheter mitral valve repair.
PhD with MicroPort CRM. A 3D model personalised from an ECG and an echocardiogram, fitted in under a second with machine learning, and compared with measurements from 91 patients to predict the response to CRT. The simulated left pre-ejection interval (LPEI) emerges as the best indicator of response.
the thesis on HAL (new tab)A reduced finite element model based on poly-affine transformations, to speed up electromechanical simulation.
STACOM 2019 paper (new tab)Electromechanical digital twins of the heart; mitral repair planning until 2026.
Numerical simulation, AI and training.
Université Côte d'Azur, supervised by Maxime Sermesant. Involved in the SimCardioTest European project.
Reduced finite element heart model based on poly-affine deformation, using SOFA in Python and C++.
Convolutional neural networks to detect soft tissue sarcoma in MRI scans.
Mathematical modelling, fluid mechanics and high performance computing.
Lycée Bellevue, Toulouse.
Digital twins of the heart: from imaging to personalised simulation.
Applied mathematics, computational engineering.
Linear equations, master's level.