2025Activity​ reportProject-TeamSIMBIOTX

3.1.1 Cells

Several‌​‌ of our applications in​​ systems medicine and systems​​​‌ biotechnology address questions at‌ the tissue micro-architecture at‌​‌ cell-and sub-cellular spatial scale.​​ In these applications we​​​‌ present each cell as‌ individual unit ("agents") in‌​‌ continuum space using mainly​​ two modeling technologies, which​​​‌ we have co-developed: center-based‌ models (CBMs) and deformable‌​‌ cell models (DCMs) 53​​, 54.

In​​​‌ CBMs, cells are parameterized‌ by a few geometric‌​‌ parameters such as the​​ cell radius, and axis​​​‌ length (e.g. to mimic‌ cell elongation prior to‌​‌ undergoing mitosis), material parameters​​​‌ and cell-kinetic parameters, and​ forces between cells are​‌ approximated as forces between​​ cell centers. CBMs have​​​‌ no explicit notion of​ shape, the volume occupied​‌ by a cell is​​ approximated by a geometric​​​‌ body (usually a sphere​ or dumb-bell) that specifies​‌ the approximate position and​​ shape. Hence despite its​​​‌ geometric representation may indicate​ a rigid cell body,​‌ the cells are usually​​ not rigid, represented by​​​‌ that their geometric representations​ can overlap depending on​‌ the forces between them.​​

In DCMs, cells are​​​‌ mimicked as deformable objects​ with an explicit representation​‌ of cell surface on​​ a mesoscopic level, usually​​​‌ by triangulation of the​ cell surfaces. The DCM​‌ can further represent cell​​ organelles. As in the​​​‌ CBM, the presented structures​ are parameterized by material​‌ parameters that are either​​ directly represented or be​​​‌ inferred from the cells'​ response on experimental perturbations.​‌ Both CBM-and DCM-cells move​​ according to force-balance equations​​​‌ that account for all​ passive forces on the​‌ cell plus active forces​​ mimicking the cell movement.​​​‌ For CBM this is​ usually one equation for​‌ a translatory cell movement,​​ while for DCM, it​​​‌ is one equation for​ each node of its​‌ triangulation. For different applications,​​ the CBM/DCM-models have to​​​‌ be adapted, which in​ particular includes the force​‌ terms in the force​​ balance equation(s) (the "equation​​​‌ of motion"). Each time,​ the model parameters have​‌ to be identified. As​​ equations of motion (force​​​‌ balance equations) are usually​ stochastic and many thousands​‌ or even millions of​​ equations have to be​​​‌ solved simultaneously, the implementation​ of those models is​‌ very time-consuming. Hence a​​ wider use of such​​​‌ models by the community​ requires access to codes​‌ that can execute those​​ models.

3.1.2 Other structures:​​​‌ networks of ellongated components​

In certain diseases collagen​‌ networks form representing architectural​​ and functional obstacles. Collagen​​​‌ bundles or fibers are​ mimicked as semiflexible chains​‌ with each node on​​ the chain being mimicked​​​‌ by a force balance​ equation as for CBMs.​‌ The same approach is​​ partially used to represent​​​‌ capillary networks as this​ permits to approximate network​‌ distortions upon physical forces​​ on the capillaries in​​​‌ a simple and computationally​ efficient way. Alternatively, vessels​‌ may be triangulated as​​ cells in the DCM.​​​‌

3.3.1 Transport‌​‌

Major fluxes considered are​​ those inside the blood​​​‌ vessels and bile conduits,‌ and between blood vessels‌​‌ or bile conduits and​​ their adjacent structures (cells,​​​‌ extracellular, extravascular space).

Currently‌ two major model types‌​‌ are used to mimic​​ transport phenomena. The first​​​‌ one are compartment models‌ where concentrations are assumed‌​‌ to be homogeneous in​​ a certain spatial compartment​​​‌ and change upon transport‌ into or from another‌​‌ compartment 5230.​​ In such models, we​​​‌ usually apply ordinary differential‌ equations (ODEs) for the‌​‌ compartment concentration as a​​ function of time. The​​​‌ second type emerges if‌ concentrations can vary in‌​‌ space (e.g. along a​​ blood vessel) in which​​​‌ case usually partial differential‌ equations (PDEs) for the‌​‌ local concentrations depending on​​ space and time are​​​‌ considered 31, 12‌. In both cases,‌​‌ the equations can be​​ derived from mass balance.​​​‌ The equations require the‌ knowledge of the flow‌​‌ rate (ODs) or local​​ flow velocity (PDE models),​​​‌ which emerge from the‌ flow models (section 3.2‌​‌).

3.3.2 Reactions

Besides​​ fluxes, the mass balance​​​‌ can be modified as‌ a consequence of chemical‌​‌ reactions. In our applications​​ modifications by chemical reactions​​​‌ mostly occur inside cells,‌ which we mostly mimic‌​‌ by ODE equations assuming​​ the number of molecules​​​‌ inside the cell is‌ sufficiently large to neglect‌​‌ stochastic fluctuations (e.g. 2​​). If the latter​​​‌ is not the case,‌ we develop master equation‌​‌ approaches to cope for​​ fluctuations. In such an​​​‌ approach, the multivariate probability‌ of a certain chemical‌​‌ species composition is tracked​​ in time, and, if​​​‌ necessary, in space by‌ subdividing the space into‌​‌ small reaction volumes (compartments)​​ much smaller than the​​​‌ cell or other local‌ volumes. The main work‌​‌ is the simulation of​​ different reaction networks that​​​‌ are believed to represent‌ alternative hypotheses on the‌​‌ reaction dynamics. The simulation​​ results are usually compared​​​‌ to experimental readout observables‌ 3.

3.4.1 Image analysis​​

Many parameters used to​​​‌ calibrate the models have‌ to be inferred from‌​‌ images 39. For​​ this purpose, the team​​​‌ has been repeatedly performing‌ image analysis. As free‌​‌ tools are usually not​​ suited for the images​​​‌ used, tools to analyze‌ images of multiple modalities‌​‌ (e.g. light sheet microscopy,​​​‌ confocal laser scanning microscopy,​ MRI) to extract information​‌ from images are developed.​​ This partially includes new​​​‌ and refined algorithms to​ better bridge the gap​‌ between experimental images and​​ computational models (e.g. 36​​​‌, 35).

For​ patients, model parameterization needs​‌ to occur from non-invasive​​ or moderately invasive modalities,​​​‌ e.g. from biomarkers or​ non-invasive imaging. While non-invasive​‌ functional imaging has been​​ a very active field​​​‌ of research, its translation​ to the clinics is​‌ impeded by a good​​ understanding of how the​​​‌ extracted parameters relate to​ the underlying tissue characteristics.​‌ A first approach consists​​ in constructing in-silico models​​​‌ of such tissue images​ and study how model​‌ parameter changes relate to​​ these in-silico images 13​​​‌. A second approach​ is to perform quantitative​‌ image analysis and correlation​​ of different image modalities​​​‌ 55. One can​ then study how non-invasive​‌ imaging, a macroscale information,​​ relates to organ microscale​​​‌ architecture, perfusion or function.​

3.4.2 Integrative, multiscale, multilevel​‌ and multicomponent models

In​​ a number of models​​​‌ the three methodology axes​ are combined to a​‌ multi-level multi-scale model (for​​ example, those aiming at​​​‌ a virtual liver at​ microscale), which raises the​‌ challenge how to choose​​ each of the model​​​‌ components and parametrise them​ (e.g. 1).

So​‌ far the mostly chosen​​ method is a systematic​​​‌ simulated parameter sensitivity analysis​ by variation of each​‌ model parameter within its​​ physiological range and studying​​​‌ how this modifies the​ agreement between model simulation​‌ result and data from​​ experiments for patients. A​​​‌ sensitivity analysis performed on​ such models would be​‌ crucial in order to​​ (i) identify the most​​​‌ significant parameters to influence​ the desired output, (ii)​‌ test the robustness of​​ a model in the​​​‌ presence of uncertainties, (iii)​ determine the interactions among​‌ parameters, and (iv) unveil​​ the optimal regions within​​​‌ the parameters space for​ optimization studies. An example​‌ is the Saltelli algorithm​​ to compute the Sobol’​​​‌ indices, a variance-based sensitivity​ analysis that exploits the​‌ variance decomposition (ANOVA) also​​ in non-linear and non-monotonic​​​‌ cases. An example of​ such sensitivity analysis applied​‌ to our virtual human​​ twins is 49.​​​‌

3.4.3 Artificial intelligence

Biophysical​ models have also been​‌ complemented by machine-learning 5​​, 48, replaced​​​‌ by 15 or mixed​ with 50, 17​‌, 22 deep-learning approaches.​​

3.4.4 Software tools

As​​​‌ mentioned above, the models​ have reached a complexity​‌ that make them difficult​​ to use or build​​​‌ upon unless they are​ accessible together with code​‌ that executes the models.​​ As a consequence an​​​‌ increasing focus is put​ on implementation of models​‌ in software. This happens​​ at the whole body​​​‌ level with the code​ "LumpedFlux" and at the​‌ tissue level with "CompuTiX".​​

4.1.1 Liver​

The objective is to​‌ establish models at multiple​​ scales and multi-scale models​​​‌ (i.e. linking intracellular functional​ units up to the​‌ whole organ scale) of​​ the different liver subsystems,​​​‌ aiming finally at a​ digital liver model (e.g.​‌ 33). After significant​​ effort in the last​​ few years, a 4th-generation​​​‌ software CompuTiX has been‌ built at the tissue‌​‌ scale that will permit​​ to execute the liver​​​‌ models at tissue level‌ and link them to‌​‌ other models at the​​ molecular level (e.g. signaling​​​‌ pathways, metabolic pathways) or‌ at the body level‌​‌ such that systematic hypothesis​​ testing with small extra​​​‌ effort should become feasible.‌ Applications in liver concern‌​‌ liver tissue architecture and​​ function in the healthy​​​‌ liver serving as a‌ reference state, as well‌​‌ as acute liver damage,​​ disease development and its​​​‌ functional consequences, as well‌ as treatment of aberrant‌​‌ states, for which a​​ prominent example is liver​​​‌ surgery. The computational models‌ integrate information from in‌​‌ vitro experiments, animal models​​ and human data. At​​​‌ the methodology level, liver‌ modeling requires all elements‌​‌ introduced in the previous​​ section, integrating agent-based modeling​​​‌ approaches for cells and‌ molecules, ODE/PDE models of‌​‌ molecular transport, flow, as​​ well as inter-cellular and​​​‌ intra-cellular reactions (which can‌ for example be signaling‌​‌ cascades, metabolic reaction networks​​ or detoxification reactions) by​​​‌ ODE models or, if‌ required, stochastic modeling methods.‌​‌

The first step is​​ to provide biologists, pharmacologists,​​​‌ toxicologists, and clinicians with‌ a better understanding of‌​‌ the interplay of the​​ many components pertaining to​​​‌ liver function, injury, and‌ the disease progression in‌​‌ a systems approach. In​​ a further step, modeling​​​‌ is increasingly used to‌ guide the design of‌​‌ experiments and data acquisition.​​ While a number of​​​‌ aims and concepts can‌ be developed based on‌​‌ animal models, where mechanisms​​ may be validated, a​​​‌ key challenge will be‌ to develop strategies and‌​‌ concepts for model and​​ parameter identification in human.​​​‌ The long-term aim is‌ to support clinicians in‌​‌ diagnosis by informing about​​ disease progression, possible disease​​​‌ origin, disease reversal, and‌ predict the possible consequences‌​‌ of therapy options. An​​ important example for a​​​‌ therapy studied in SIMBIOTX‌ is liver surgery 4‌​‌.

Liver disease partially​​ impacts on other organs​​​‌ such as heart, kidney‌ and lung, which might‌​‌ therefore be addressed if​​ required by the clinical​​​‌ questions.

4.1.2 Congenital heart‌ disease

Congenital Heart Disease‌​‌ (CHD) consists in diseases​​ that affect children born​​​‌ with heart or connecting‌ large vessel abnormalities. Pulmonary‌​‌ hypertension is a disease​​ that has several etiologies,​​​‌ one of which is‌ CHD.

While great advances‌​‌ have been made in​​ the last decades in​​​‌ their clinical treatment (mainly‌ through surgery), these patients‌​‌ still suffer from significant​​ mortality and morbidity, due​​​‌ in part to interactions‌ between heart, systemic circulation,‌​‌ pulmonary circulation and other​​ components such as implanted​​​‌ graft or devices. The‌ goal here is to‌​‌ perform patient-specific modeling to​​ better understand such interactions​​​‌ (e.g. 45). Choosing‌ the treatment option (surgical,‌​‌ interventional, drug) and optimizing​​ it based on modeling​​​‌ opens up several research‌ directions.

4.1.3 In vitro‌​‌ cell populations, tumors and​​ cancer

A tumor can​​​‌ be malign (cancerous) or‌ benign. A malign tumor‌​‌ can grow and spread​​ to other parts of​​​‌ the body. A benign‌ tumor can grow but‌​‌ will not spread. Modeling​​​‌ of growing cell populations,​ early tumor growth, different​‌ phases in tumor development​​ (e.g. invasion and intravasation),​​​‌ have hence been a​ regular work subject of​‌ SIMBIOTX members as it​​ does not only provide​​​‌ interesting insight into the​ biological processes underlying cancer​‌ development, but also permits​​ to study and develop​​​‌ the modeling concepts and​ methodology. Many cell-mechanisms are​‌ first studied in-depth in​​ in vitro cell populations​​​‌ such as the effect​ of mechanical stress on​‌ cell growth and proliferation​​ 10, which makes​​​‌ them prone to be​ implemented first in models​‌ of the in vitro​​ setting before integrating them​​​‌ into in vivo tumor​ growth models.

8.1.1 Modeling of liver​​ disease progression from liver​​​‌ steatosis to cirrhosis and‌ HCC: digital twins

Participants:‌​‌ Matteo Pedrazzi, Jieling​​ Zhao, Dirk Drasdo​​​‌.

Metabolic dysfunction-associated steatotic‌ liver disease (MASLD), formerly‌​‌ known as non-alcoholic fatty​​ liver disease (NAFLD), is​​​‌ a chronic liver disorder‌ characterized by excessive lipid‌​‌ accumulation in hepatocytes. This​​ leads to hepatic steatosis​​​‌ and may progress to‌ more severe conditions. MASLD‌​‌ is responsible for approximately​​ one in every 25​​​‌ deaths worldwide each year,‌ with advanced stages including‌​‌ cirrhosis and hepatocellular carcinoma​​ (HCC). Disease progression is​​​‌ a very complex process‌ due to its multiscale‌​‌ nature, involving intracellular, intercellular,​​ tissue-level, and body level​​​‌ processes that evolve over‌ many years. The gold‌​‌ standard for staging of​​ the disease is histology,​​​‌ requiring invasive taking of‌ biopsies,but in order to‌​‌ avoid taking biopsies, diagnosis​​ and staging in clinics​​​‌ is often based on‌ combining non-invasive methods such‌​‌ as fibroscans and blood​​ serum analysis. However, biopsies​​​‌ and alternative methodologies often‌ suggest different disease stages‌​‌ indicating a lack of​​ understanding of how these​​​‌ different modalities are related‌ to each other. This‌​‌ motivated us to develop​​ a disease-progression model integrating​​​‌ factors that permits to‌ study the different aspects‌​‌ simultaneously in one model.​​ The starting points are​​​‌ formed by the agent-based‌ model in spatial liver‌​‌ lobule microarchitecture previously developed​​ in our group 56​​​‌, and more refined‌ cell-cell crosstalk model addressing‌​‌ a potential strategy to​​​‌ slow down fibrosis 20​. In the latter​‌ works, a fibrosis progression​​ and alteration by ECM1​​​‌ as a potential therapeutic​ target has been studied​‌ in wild type and​​ ECM1-knockout mice, while a​​​‌ compartment model has been​ developed to confirm consistency​‌ of the found effects​​ of ECM1 on the​​​‌ known or hypothesized components​ of the liver fibrosis​‌ network. The compartment model​​ assumes a well-stirred liver​​​‌ tissue and is able​ to explain the qualitative​‌ tendencies but not the​​ quantitative behavior of the​​​‌ considered species in space​ and time. For this,​‌ a spatial-temporal model at​​ liver micro-architecture is indispensable.​​​‌

We started to newly​ implement the tissue components​‌ in the `CompuTiX` software​​ which is very time​​​‌ consuming as the community-tool​ friendly code architecture requires​‌ coding of each model​​ from scratch. This will​​​‌ be followed by integration​ of relevant models for​‌ metabolism, cell-cell-and intracellular signaling​​ such as ammonia detoxification​​​‌ and signaling models.

Those​ new components involve an​‌ intracellular model of lipid​​ and urea metabolism. The​​​‌ urea cycle has already​ been parameterized for mice​‌ earlier [Ghallab et al.,​​ 2016], while in 2025​​​‌ this model has be​ extended by incorporating transaminases,​‌ which serve as indicators​​ of liver function in​​​‌ clinics.

Moreover, we worked​ out a computational pipeline​‌ to integrate transcriptomics data​​ from the GeoMX platform​​​‌ of our collaborators building​ on the work of​‌ Montagud et al. 44​​. The core idea​​​‌ is to construct a​ gene regulatory network (GRN)​‌ and perform stochastic simulations​​ to map microenvironmental cellular​​​‌ inputs to phenotypic outputs.​ This approach enables the​‌ investigation of alternative drivers​​ of cellular behavior, including​​​‌ apoptosis, migration, uncontrolled proliferation,​ and the production and​‌ secretion of molecular signals,​​ permitting to infer possible​​​‌ cell fate developments in​ disease progression.

Collaborators: S.​‌ Hammad, S. Dooley (Heidelberg​​ University), M. Cascante (Barcelona​​​‌ University), Ursula Klingmüller (German​ Cancer Center Heidelberg), Jens​‌ Timmer (University of Freiburg),​​ Heike Bantel (University Clinics​​​‌ Hannover) and others.

8.1.2​ Modeling liver the generation​‌ of fibrotic scar formation​​

Participants: Jieling Zhao,​​​‌ Dirk Drasdo.

We​ extended and further refined​‌ our digital liver twin​​ model of the formation​​​‌ of fibrotic scars in​ septal fibrosis and biliary​‌ fibrosis. While in septal​​ fibrosis the scars are​​​‌ sharply located at the​ connecting line between central​‌ veins of neighboring liver​​ lobules in biliary fibrosis,​​​‌ they are scattered at​ the portally located borders​‌ between neighboring lobules. We​​ directly compare our simulation​​​‌ results to data in​ animal models and human​‌ patient data. The model​​ is able to explain​​​‌ critical differences between biliary​ fibrosis pattern in mice​‌ and human, and the​​ septal fibrosis pattern in​​​‌ mice after repetitive overdosing​ of drugs causing cell​‌ death of Cytochrom 450-expressing​​ hepatocytes. In human one​​​‌ would expect this situation​ to occur after repetitive​‌ overdosing of acetaminophen (paracetamol),​​ for which histology data​​​‌ is very sparse.

Main​ collaborators: S. Hammad, S.​‌ Dooley (Heidelberg University), J.G.​​ Hengstler, IfADo, Dortmund.

8.1.3​​​‌ Multilevel modeling of flow​ and transport in liver​‌ lobules in health and​​ disease

Participants: Peter Kottman​​, Dirk Drasdo,​​​‌ Irene Vignon-Clementel.

The‌ liver guarantees its function‌​‌ as main detoxifying organ​​ by a complex microarchitecture​​​‌ that maximizes the interface‌ area for exchange of‌​‌ metabolites between blood and​​ hepatocytes, whereby some molecular​​​‌ species are secreted into‌ the biliary system 12‌​‌. Most pharmacokinetic and​​ pharmacodynamic models ignore the​​​‌ microarchitecture and instead lump‌ the liver lobules, which‌​‌ constitute the minimal repetitive​​ functional and anatomical liver​​​‌ tissue units, into one‌ or two well-mixed compartments.‌​‌ However, how critical the​​ microarchitecture is for liver​​​‌ function is largely unclear.‌ This is also reflected‌​‌ in computational models of​​ liver metabolism or liver​​​‌ toxicology. Most of these‌ make the assumption of‌​‌ a well-mixed liver compartment​​ or well-mixed liver subcompartments​​​‌ but a systematic study‌ under which conditions the‌​‌ well-mixed assumption applies or,​​ if it is possible​​​‌ to correct the deviations‌ of well-mixed models from‌​‌ their spatially resolved counterparts,​​ is missing.

To explore​​​‌ the impact of microarchitecture‌ on liver function at‌​‌ higher levels of resolution,​​ several upscaling methods were​​​‌ studied. On one hand,‌ a theoretical correspondence was‌​‌ established between the lobular​​ flow and transport model​​​‌ in spatially resolved networks‌ and a homogenized porous‌​‌ media flow and transport​​ model to bridge the​​​‌ information gap between microscale‌ and larger tissue- and‌​‌ organ-scale models of liver​​ function. On the other​​​‌ hand, compartment models of‌ liver clearance were studied.‌​‌ Based on the earlier​​ ideas formulated in ref.​​​‌ ([CellierePhDThesis]), a‌ strategy for multilevel analysis‌​‌ of liver clearance was​​ devised, where the appropriate​​​‌ choice of resolution and‌ compartment model complexity is‌​‌ guided by the scale​​ of interest and the​​​‌ scale at which data‌ is available. Such models‌​‌ were confronted with simulations​​ in fully resolved microarchitectures​​​‌ (see above), and can‌ be seen as a‌​‌ bridging point between the​​ spatially resolved models that​​​‌ represent each cell individually‌ and well-mixed single-compartment models‌​‌ that disregard completely the​​ spatial heterogeneity of liver.​​​‌

Collaborators: E. Rohan, UWB‌ Pilsen, Czech Republic

8.1.4‌​‌ Temporal diffusion spectroscopy for​​ the characterization of MASH​​​‌

Participants: Charles Boulitrop,‌ Jiří Pešek, Dirk‌​‌ Drasdo.

The gold​​ standard in the staging​​​‌ of many liver diseases‌ is the evaluation of‌​‌ a histological sample of​​ the patient. Obtaining tissue​​​‌ material for histopathological evaluation‌ requires invasive interventions such‌​‌ as taking biopsies, which​​ underlies strict criteria due​​​‌ to the possible adverse‌ effects. The objective is‌​‌ hence to gather knowledge​​ on histopathological characteristics non-invasively,​​​‌ as this not only‌ minimizes size effects but‌​‌ also permits longitudinal follow-ups.​​ A typical use case​​​‌ is the classification of‌ metabolic dysfunction-associated steatohepatitis (MASH),‌​‌ which includes hepatocyte ballooning.​​ Ballooning is characterized by​​​‌ accumulation of lysis vesicles‌ in the cytoplasmic space,‌​‌ hepatocyte swelling and alterations​​ of the intermediate filament​​​‌ cytoskeleton. Temporal diffusion spectroscopy‌ (TDS), a development of‌​‌ Diffusion-Weighted-Magnetic Resonance Imaging (DW-MRI),​​ has recently enabled identification​​​‌ of apparent diffusion coefficients‌ that may permit to‌​‌ probe ballooning non-invasively.

Using​​ the team's novel simulation​​​‌ software CompuTiX, agent-based models‌ simulating the DW-MRI process‌​‌ have been developed. The​​​‌ main physical processes at​ stake in DW-MRI, diffusion​‌ of magnetic moments and​​ magnetic precession, were simulated​​​‌ in CompuTiX and have​ been verified against the​‌ conventional continuum-based approach (namely​​ finite-element method), with good​​​‌ agreement. Additionally, models of​ DW-MRI in a constrained​‌ synthetic microarchitecture have been​​ developed in CompuTiX, including​​​‌ specific gradient pulse sequences​ used for in vitro​‌ experiments with "phantoms". Apparent​​ diffusion coefficients are then​​​‌ derived from the simulation​ results. These display the​‌ expected behavior with regard​​ to theory. The final​​​‌ goal is to study​ how changes in the​‌ liver tissue microarchitecture may​​ impact on MRI-signals.

In​​​‌ 2025 we first performed​ extensive simulations with the​‌ model and some model​​ refinements for phantoms and​​​‌ compared the numerical results​ of apparent diffusion coefficients​‌ with those from experiments.​​ Eventually, we found good​​​‌ agreement between experimental and​ theoretical results. In a​‌ next step we adapted​​ the model to permit​​​‌ proof-of-concept simulations in a​ digital twin of the​‌ liver tissue thought. This​​ should constitute the final​​​‌ step.

Collaborators: P. Garteiser,​ B. Van Beers (Inserm)​‌

8.1.5 A digital liver​​ twin to demonstrate the​​​‌ significance of disease-related remodeling​ of liver architecture in​‌ ammonia detoxification

Participants: Jules​​ Dichamp, Dirk Drasdo​​​‌, Geraldine Cellière,​ Noémie Boissier.

Hyperammonemia​‌ i.e., a critically elevated​​ ammonia blood concentration, can​​​‌ lead to encephalopathy and​ patients' death and constitutes​‌ the major reason for​​ acute liver failure. We​​​‌ introduce a digital liver​ twin that reflects the​‌ remodeling of architecture and​​ all key processes relevant​​​‌ in ammonia detoxification during​ fibrosis development. We demonstrate,​‌ that the architectural changes​​ alone can explain most​​​‌ of the experimentally observed​ changes in ammonia detoxification​‌ during fibrosis. Our findings​​ suggest a novel modeling​​​‌ strategy in toxicology modeling​ in tissues during acute​‌ liver tissue damage or​​ disease development, by first​​​‌ studying the effect of​ tissue remodeling of a​‌ toxic substance, and subsequently​​ adapt intracellular processes to​​​‌ capture the observed concentrations​ of toxic substances.

In​‌ 2025 we performed simulations​​ based on new fibrosis​​​‌ fraction assessments during the​ 12 months-lasting disease progression​‌ process indicating that by​​ far the largest impact​​​‌ on fibrosis progression occurs​ in the last 2​‌ months i.e., between month​​ 10 and 12. The​​​‌ paper draft is now​ about to be finalized.​‌

Collaborators: J.G. Hengstler, A.​​ Ghallab and coworkers from​​​‌ Leibnitz Institute IFADO, Dortmund,​ Germany.

8.2.1 3D​​​‌ digital histopathology: a new​ methodology for morphological characterization​‌ of the human liver​​

Participants: Dirk Drasdo,​​​‌ Mathieu de Langlard,​ Tobias Schnirer, Irene​‌ Vignon-Clementel.

2D histopathology​​ remains the gold standard​​​‌ for diagnosing and studying​ organ diseases, but the​‌ complex 3D tissue organization​​ demands more spatially resolved​​​‌ and consistent representations for​ improved understanding and early​‌ diagnosis. In 26 we​​ propose a novel automatic​​​‌ digital pathology workflow for​ 3D histological reconstruction of​‌ liver tissue, combining 2D​​ whole-slide serial image registration,​​​‌ shown to outperform high-end​ commercial software, and segmentation​‌ of key hepatic structures​​ from meso- to microscale.​​ Particular attention is given​​​‌ to the fine segmentation‌ of hepatic trees, proposing‌​‌ a method to distinguish​​ disconnected portal and central​​​‌ vein structures based solely‌ on bile duct staining.‌​‌ Furthermore, we achieve fully​​ automatic 3D liver lobule​​​‌ segmentation using a refined‌ watershed algorithm and a‌​‌ central vein (CV) seeding​​ strategy based on CV-specific​​​‌ morphological and topological features.‌ This enables the analysis‌​‌ of large and representative​​ liver samples across scales.​​​‌ Morphological quantification derived from‌ these segmentations provides new‌​‌ insights into the 3D​​ microscale organization of the​​​‌ liver, including an algorithm‌ to topologically connect microscale‌​‌ subtrees to higher-scale trees.​​ This allows for the​​​‌ analysis of hepatic tree‌ topology at an unprecedented‌​‌ level of detail and​​ the 3D structural characterization​​​‌ of lobular and interlobular‌ regions. The resulting segmentations‌​‌ enable computational models to​​ assess the functional impact​​​‌ of architectural changes. The‌ full digital workflow will‌​‌ be released as fully​​ documented, open-source code to​​​‌ support adoption in biomedical‌ research and potentially clinical‌​‌ settings. While this study​​ focuses on the liver,​​​‌ the methodology is applicable‌ to other organs for‌​‌ digital pathology.

As specific​​ usecase the pipeline permitted​​​‌ us to characterize lobule‌ sizes (in 3D) the‌​‌ orientation. of sinusoids within​​ a human lobule, radii​​​‌ and segment length of‌ the portal and central‌​‌ vene vascular trees etc.​​

Collaborators: APHP (Hôpital Paul​​​‌ Brousse, Le Kremlin-Bicetre), INSERM‌ U1193

8.3.1‌​‌ Geometric and Hemodynamic Study​​ of the Portal Vein:​​​‌ Identification of Predictive Factors‌ for Portal Thrombosis after‌​‌ Extended Liver Resection

Participants:​​ Morgane Garreau, Ana​​​‌ Vlasceanu, Amaury Facque‌, Weiqiang Liu,‌​‌ Irene Vignon-Clementel.

Hepatocellular​​ carcinoma, cholangiocarcinoma and liver​​​‌ metastases of colorectal cancer‌ account for the vast‌​‌ majority of liver tumor​​ lesions. Cholangiocarcinoma (CK) is​​​‌ the most common biliary‌ malignancy. It accounts for‌​‌ 3% of all digestive​​ tumors, with an incidence​​​‌ of 5,000 new cases‌ per year in France.‌​‌ Apart from liver transplantation,​​ which is currently not​​​‌ an option for the‌ majority of patients (organ‌​‌ shortage, risk of recurrence),​​ surgical resection is considered​​​‌ the main curative treatment.‌ Long-term survival is between‌​‌ 3 and 6 months​​ without resection, 12 months​​​‌ with chemotherapy alone, 24‌ months with resection and‌​‌ positive margins, and between​​ 36 and 48 months​​​‌ with resection and negative‌ margins. In most surgical‌​‌ series, 5-year survival is​​ between 10 and 40%.​​​‌

In this work, extended‌ liver resection for perihilar‌​‌ CK is considered, as​​ these tumors require complex​​​‌ resections associated with high‌ morbidity/mortality and, in particular,‌​‌ high risk of portal​​ vein thrombosis. This complication​​​‌ occurs in approximately 10%‌ of patients and can‌​‌ be fatal. 3 groups​​ of patients are investigated:​​​‌ patients who underwent portal‌ resection (PR) with consecutive‌​‌ thrombus formation, patients who​​​‌ underwent PR but without​ thrombus development, and patients​‌ who did not undergo​​ PR and did not​​​‌ develop a portal thrombus.​ Their portal vein geometries​‌ are reconstructed on the​​ basis of postoperative CT​​​‌ scans. A fourth group​ is built based on​‌ preoperative CT scans of​​ patients who had PR​​​‌ and developed thrombus, where​ surgery is replaced in​‌ silico by a virtual​​ graft.

Geometric analysis and​​​‌ CFD simulations were conducted​ in 2024. This work​‌ has continued with a​​ specific focus on shear​​​‌ and derived hemodynamic quantities,​ which have been reported​‌ to be pro-thrombotic. These​​ results have been presented​​​‌ as an oral communication​ in an international conference.​‌ A collaboration has been​​ established with LadHyX to​​​‌ model the thrombus formation​ in 2D.

Collaborators: N.​‌ Golse, E. Vibert at​​ AP-HP - Hôpital Paul​​​‌ Brousse; L. Sala at​ INRAE; G. Cardillo, A.​‌ Barakat at LadHyX -​​ Ecole Polytechnique

8.3.2 Hemodynamics​​​‌ modeling for liver surgery:​ digital twins

Participants: Francesco​‌ Songia, Roel Meiburg​​, Kevin Hakkakian,​​​‌ Ramdane Bessaïd, Clémence​ Finotto, Irene Vignon-Clementel​‌.

To evaluate the​​ risk of portal hypertension​​​‌ after partial hepatectomy, our​ team proposed a lumped-parameter​‌ model to predict postoperative​​ hemodynamic changes. This year,​​​‌ our clinical cohort was​ expanded with 30 additional​‌ patients. On the modeling​​ side, we focused on​​​‌ incorporating perioperative events and,​ in particular, on characterizing​‌ how vascular resistances in​​ the lumped model evolve​​​‌ from the pre- to​ the post-resection state. Our​‌ goal is to relate​​ these resistance changes to​​​‌ specific patient-related factors, thus​ capturing how individual patients​‌ may respond differently to​​ liver resection. To this​​​‌ end, we use data​ describing patients' medical history,​‌ underlying disease, and perioperative​​ events during surgery (including​​​‌ injected volumes, blood loss,​ anesthesia, and others). In​‌ particular, we also focused​​ on patients who underwent​​​‌ portal vein embolization before​ the resection. Finally, further​‌ sensitivity and identifiability analyses​​ demonstrated that time-series data​​​‌ (which are technically and​ clinically feasible) of portal​‌ pressure and flow pre-intervention​​ potentially contain information on​​​‌ blood pressure inside the​ liver tissue, and may​‌ hence contain information on​​ whether any obstruction is​​​‌ pre- or post-sinusoidal.

Collaborators:​ N. Golse, S. Roullet,​‌ A. Coilly, and E.​​ Vibert at APHP -​​​‌ Hôpital Paul Brousse

8.3.3​ Numerical investigation of particle​‌ aggregate steering with magnetic​​ resonance navigation for targeted​​​‌ embolization

Participants: Mahdi Rezaei​ Adariani, Jiří Pešek​‌, Irene Vignon-Clementel.​​

Magnetic resonance navigation (MRN)​​​‌ uses the strong magnetic​ field of clinical MRI​‌ scanners to steer magnetic​​ microparticles toward tumors for​​​‌ localized drug delivery. A​ key mechanism is particle​‌ aggregation, which arises from​​ magnetic dipole interactions in​​​‌ the homogeneous MRI field.​ Aggregation increases local drug​‌ concentration and alters the​​ force balance acting on​​​‌ the particles. Despite these​ advantages, accurate control of​‌ aggregate transport remains challenging​​ in vascular environments.

Several​​​‌ coupled forces govern aggregate​ motion. Magnetic, hydrodynamic, gravitational,​‌ and wall-induced forces interact​​ in confined, flowing vessels.​​​‌ These interactions complicate steering,​ especially near vessel walls​‌ and under pulsatile blood​​ flow. As a result,​​ aggregate trajectories often deviate​​​‌ from predictions based on‌ simplified models. In addition,‌​‌ repeated injections may modify​​ local hemodynamics through partial​​​‌ downstream blockage, which further‌ affects steering accuracy. Therefore,‌​‌ before translating MRN to​​ clinical applications, a comprehensive​​​‌ study of aggregate motion‌ under physiological conditions is‌​‌ required to improve predictive​​ accuracy.

This project is​​​‌ organized into several stages.‌ First, we developed a‌​‌ numerical framework for chain-like​​ aggregates, which represent an​​​‌ ideal aggregate shape under‌ a uniform magnetic field.‌​‌ The model accounts for​​ the dominant forces acting​​​‌ during MRN, including drag,‌ magnetic forces, gravitational forces,‌​‌ and near-wall effects. We​​ validated this framework using​​​‌ a series of in‌ vitro experiments conducted under‌​‌ MRN-relevant conditions.

We currently​​ focus on realistic blood​​​‌ flow modeling. We compare‌ numerical predictions with 4D‌​‌ flow MRI data from​​ in vivo experiments conducted​​​‌ at the CHUM hospital‌ research center. This step‌​‌ is essential because blood​​ flow serves as the​​​‌ primary transport mechanism for‌ aggregates in hepatic arteries.‌​‌

This approach links particle-scale​​ physics with clinically realistic​​​‌ flow environments and may‌ improve the reliability of‌​‌ MRN-guided drug delivery.

Collaborators:​​ Gilles Soulez (CR-CHUM, Montreal,​​​‌ Canada), Charlotte Debbaut (group‌ bioMMeda, UGent, Belgium)

8.3.4‌​‌ Computational modeling for TIPS​​ shunt implantation and pre-procedural​​​‌ planning

Participants: Pavlos Varsos‌, Friederike Schäfer,‌​‌ Irene Vignon-Clementel.

The​​ Transjugular Intrahepatic Portosystemic Shunt​​​‌ (TIPS) is a well-established‌ therapeutic procedure for patients‌​‌ with liver cirrhosis, shown​​ to improve transplant-free survival.​​​‌ It consists of a‌ stented graft, typically made‌​‌ of PTFE, that connects​​ the portal vein to​​​‌ a hepatic vein within‌ the liver parenchyma, thereby‌​‌ reducing the portal pressure​​ gradient. However, excessive shunting​​​‌ through larger graft diameters‌ can increase systemic ammonia‌​‌ levels, predisposing patients to​​ hepatic encephalopathy (HE). Computational​​​‌ modeling can therefore serve‌ as a valuable predictive‌​‌ tool for patient selection​​ and preoperative planning, optimizing​​​‌ shunt diameter and assessing‌ the portosystemic pressure gradient.‌​‌ In this study, patient-specific​​ TIPS geometries are segmented​​​‌ from CT images and‌ analyzed using computational fluid‌​‌ dynamics (CFD) with realistic​​ boundary conditions to investigate​​​‌ the influence of geometrical‌ and flow-related factors. The‌​‌ analysis includes variations in​​ shunt positioning (right, left,​​​‌ and main portal vein)‌ and diameter (8, 9,‌​‌ and 10 mm). Flow​​ investigations assess different inflow​​​‌ and outflow configurations.

Collaborators:‌ N. Golse, APHP -‌​‌ Hôpital Paul Brousse

8.3.5​​ Whole-body hemodynamics of cirrhotic​​​‌ liver patients

Participants: Friederike‌ Schäfer, Irene Vignon-Clementel‌​‌.

Liver cirrhosis leads​​ to changes in the​​​‌ hemodynamics not only within‌ the liver but also‌​‌ in the portal venous​​ system, the splanchnic circulation,​​​‌ overall systemic hemodynamics. Further,‌ cardiac function can be‌​‌ affected by cirrhosis. Hepatic​​ surgical interventions like TIPS​​​‌ placement or liver transplantation‌ alter further whole-body hemodynamics.‌​‌ We started to build​​ a 0-D model of​​​‌ the whole-body circulation to‌ investigate the effects of‌​‌ these interventions before they​​ are performed. The baseline​​​‌ model includes only the‌ heart and liver, where‌​‌ the single fiber model​​ (SFM) has been used​​​‌ as a representation of‌ the heart chambers. This‌​‌ allows to model cardiomyopathy​​​‌ due to cirrhosis development​ or occurring after TIPS/transplantation.​‌ The group previously showed​​ that the SFM has​​​‌ a greater physiological interpretability,​ which will be important​‌ for clinical applications, and​​ is better suited for​​​‌ personalized simulations compared to​ the more frequently used​‌ time varying elastance model​​ 16. Following discussions​​​‌ with clinicians, the model​ will further incorporate relevant​‌ organs influenced by portal​​ hypertension, specifically the digestive​​​‌ organs, spleen, and kidneys.​ This model is currently​‌ being implemented in the​​ new 0D-code called LumpedFlux.​​​‌

8.4.1 Myocardial perfusion simulation​‌ for coronary artery disease:​​ combination of Machine Learning​​​‌ and physical simulation

Participants:​ Raoul Sallé de Chou​‌, Francesco Songia,​​ Tobias Schnirer, Irene​​​‌ Vignon-Clementel.

Blood perfusion​ imaging is a challenging​‌ and expensive imaging modality​​ for the analysis of​​​‌ myocardial perfusion and the​ diagnosis of coronary artery​‌ diseases (CAD). A previous​​ model was developed for​​​‌ myocardial perfusion simulation for​ coronary artery disease in​‌ 46 to replace the​​ actual imaging exam with​​​‌ a numerical twin and​ conduct it via simulations.​‌ The model aims at​​ reproducing [15O]H2O PET imaging​​​‌ exam using only CT​ scans as input. However,​‌ in addition to a​​ high computational cost, the​​​‌ simulation fails to accurately​ reproduce some diseases, particularly​‌ those that affect medium-size​​ coronary branches. The main​​​‌ goal of this project​ is to combine Machine​‌ Learning (ML) methods with​​ physical simulations in order​​​‌ to improve the current​ simulation pipeline while diminishing​‌ the dependency on patient​​ data for the ML​​​‌ models. To achieve this,​ each part of the​‌ simulation is to be​​ replaced by an ML​​​‌ model.

The perfusion model​ integrates a 1D Navier-Stokes​‌ (NS) model for the​​ coronary arteries coupled with​​​‌ a Darcy model to​ simulate perfusion within the​‌ myocardial volume. To address​​ the myocardium component, we​​​‌ developed a finite volume​ informed Graph Neural Network​‌ (GNN) 50. To​​ predict the solution of​​​‌ the 1D NS equations,​ a transformer encoder model​‌ was employed to estimate​​ the pressure distribution from​​​‌ a given flow distribution​ within the coronary network.​‌

The second key component​​ of the simulation involves​​​‌ the generation of synthetic​ vascular trees based on​‌ the “Constructive Constrained Optimization”​​ (CCO) method 51.​​​‌ However, this method method​ has notable limitations, including​‌ high computational costs, a​​ tendency to produce sub-optimal​​ solutions. To address these​​​‌ limitations, our work focuses‌ on developing more optimal‌​‌ synthetic trees by generating​​ them directly from terminal​​​‌ points uniformly sampled within‌ the myocardial volume. We‌​‌ developed a gradient-based method​​ to construct optimized binary​​​‌ trees. In our new‌ approach, instead of minimizing‌​‌ total volume, our approach​​ minimizes a transport cost​​​‌ 51. Preliminary results‌ indicate that this method‌​‌ generates trees more efficiently​​ and achieves more globally​​​‌ optimal solutions than the‌ CCO method. Additionally, this‌​‌ approach paves the way​​ for integrating deep learning​​​‌ methods into tree generation.‌

Currently, the simulation relies‌​‌ on an estimate of​​ the total blood flow​​​‌ in the coronary arteries‌ for a healthy patient,‌​‌ derived from myocardial mass.​​ A more accurate estimation​​​‌ of a patient's total‌ blood flow can be‌​‌ obtained using PET images,​​ which provide insights into​​​‌ the state of blood‌ flow in the microvascular‌​‌ network. As a final​​ component of the project​​​‌ (23), we‌ trained a Machine Learning‌​‌ model to predict total​​ blood flow as measured​​​‌ by PET images. While‌ the model improved total‌​‌ flow prediction on a​​ test set compared to​​​‌ the mass-based estimate, further‌ investigation was undertaken to‌​‌ identify better features derived​​ from coronary artery geometries​​​‌ and their relationship to‌ CAD using GNN.

Collaborators:‌​‌ L. Najman (ESIEE -​​ U Gustave Eiffel), H.​​​‌ Talbot (CentraleSupelec, INRIA OPIS),‌ L. Papamanolis (Stanford university,‌​‌ USA, California) Heartflow inc.​​ (USA, California).

8.4.2 Potts​​​‌ Shunt as a palliative‌ treatment option for suprasystemic‌​‌ idiopathic Pulmonary Arterial Hypertension:​​ an in-silico modelling study​​​‌

Participants: Pavlos Varsos,‌ Irene Vignon-Clementel.

This‌​‌ project focuses on improving​​ surgical decision making for​​​‌ pediatric patients with severe‌ drug resistant pulmonary arterial‌​‌ hypertension through patient specific​​ computational modelling of the​​​‌ Potts shunt procedure. By‌ integrating detailed three dimensional‌​‌ simulations of the shunt​​ region with a closed​​​‌ loop lumped parameter cardiovascular‌ model, we evaluate both‌​‌ local hemodynamics and global​​ cardiac function following shunt​​​‌ implantation. The framework enables‌ comparison of alternative surgical‌​‌ strategies. In close collaboration​​ with leading clinical centers​​​‌ in France and the‌ United States, the model‌​‌ is calibrated and and​​ aimed to be validated​​​‌ using pre and post‌ operative data, and complemented‌​‌ by sensitivity and uncertainty​​ analyses using reduced order​​​‌ (0D) models. Overall, this‌ work aims to identify‌​‌ key physiological and surgical​​ factors governing procedural success,​​​‌ support patient stratification, and‌ lay the foundation for‌​‌ model informed planning of​​ Potts shunt interventions as​​​‌ a viable alternative to‌ lung transplantation in selected‌​‌ patients.

Collaborators: S. Pant,​​ A. Marsden (Stanford University,​​​‌ USA), J. Feinstein &‌ G. Adamson (Lucile Packard‌​‌ Children's Stanford University Hospital,​​ USA), M. Grady (St.​​​‌ Louis Children's Hospital, USA),‌ E. Valdeolmillos & S.‌​‌ Hascoet, (Hôpital Marie Lannelongue,​​ France)

8.4.3 Reduced Order​​​‌ Modelling for the circle‌ of Willis

Participants: Garance‌​‌ Martin, Pavlos Varsos​​, Irene Vignon-Clementel.​​​‌

This project focuses on‌ neonatal clinical cases requiring‌​‌ ECMO as a life​​ saving intervention for refractory​​​‌ cardiac or pulmonary failure.‌ In this population, cannulation‌​‌ is typically performed via​​​‌ the right common carotid​ artery and right internal​‌ jugular vein. A key​​ clinical challenge arises during​​​‌ decannulation, when surgeons must​ choose between carotid ligation​‌ or vascular reconstruction. While​​ ligation is faster and​​​‌ technically simpler, it permanently​ sacrifices the right carotid​‌ artery and may compromise​​ cerebral perfusion, particularly in​​​‌ patients with an incomplete​ Circle of Willis, increasing​‌ the risk of neurological​​ injury. Reconstruction preserves arterial​​​‌ continuity and bilateral cerebral​ flow but carries higher​‌ surgical complexity and complication​​ risk. To address this​​​‌ trade off, the objective​ of this project is​‌ to develop a patient​​ specific 0D hemodynamic model​​​‌ that enables digital twin​ simulations of cannulation and​‌ decannulation strategies and predicts​​ their impact on cerebral​​​‌ and collateral flow distribution.​

Collaborators: AP-HP Hôpital Armand​‌ Trousseau

8.4.4 Impact of​​ hemodynamics shear forces on​​​‌ sickle cell disease-related cerebral​ vasculopathy development

Participants: Morgane​‌ Garreau, Weiqiang Liu​​, Lazaros Papamanolis,​​​‌ Irene Vignon-Clementel.

Sickle​ cell disease (SCD) is​‌ the most common inherited​​ blood disorder in the​​​‌ world. It is associated​ with serious complications such​‌ as cerebral vasculopathy (CV).​​ CV is characterized by​​​‌ remodeling of the intracranial​ carotid arteries (ICA), resulting​‌ in stenosis and occlusion​​ that can lead to​​​‌ stroke, most commonly in​ young patients aged 2​‌ to 5 years. A​​ proven method for screening​​​‌ patients at high risk​ for stroke is transcranial​‌ Doppler. A time-averaged maximum​​ velocity (TAMV) greater than​​​‌ 200 cm/s has been​ proposed as an effective​‌ threshold to identify patients​​ with the highest risk.​​​‌ However, the causes of​ these pathological blood flow​‌ velocities remain unclear.

A​​ retrospective cohort of 15​​​‌ patients has been investigated:​ 5 patients under the​‌ age of 5 years,​​ 5 between the age​​​‌ of 5 and 18​ years, and 5 adults.​‌ CFD simulations have been​​ conducted and differences in​​​‌ terms of geometries and​ hemodynamics, in particular shear​‌ rate and wall shear​​ stress 21, have​​​‌ been found between the​ groups. This is the​‌ first paper on that​​ subject, and has thus​​​‌ been the subject of​ a commentary in the​‌ American J. of Hematology​​ 47.

However, after​​​‌ medical reexamination, none of​ the patients develops a​‌ vasculopathy, reducing the impact​​ of these outcomes. Patients​​​‌ who developed vasculopathy despite​ treatment and have longitudinal​‌ medical data are currently​​ being recruited to bridge​​​‌ this gap.

From another​ perspective, this project is​‌ part of a collaboration​​ with biologists at IMRB.​​​‌ 3D printed models have​ been created based on​‌ the patient-specific geometries segmented​​ from CT scans and​​​‌ used for CFD simulations​ 34. The 3D​‌ printed geometries have been​​ seeded with endothelial cells​​​‌ and set under different​ flow conditions 24.​‌

Collaborators: Saskia Eckert (PhD​​ student, ENSAM, IMRB, EFS,​​​‌ Créteil), Kim-Anh Nguyen-Peyre (PhD,​ IMRB, EFS, Créteil), Suzanne​‌ Verlhac (MD, Hôpital Robert​​ Debré AP-HP), Pablo Bartolucci​​​‌ (MD, Univ Paris Est​ Créteil/Hôpitaux Universitaires Henri Mondor​‌ AP-HP, Créteil)

8.4.5 Hemodynamics​​ impact of the TEVAR​​​‌ procedure based on 4D​ flow MRI

Participants: Morgane​‌ Garreau, Ramdane Bessaid​​, Irene Vignon-Clementel.​​

Thoracic EndoVascular Aortic Repair​​​‌ (TEVAR) is a minimally‌ invasive procedure developed in‌​‌ the 1990s and the​​ current gold standard to​​​‌ treat main aortic pathologies‌ such as thoracic aortic‌​‌ aneurysm and type B​​ aortic dissection. All these​​​‌ pathologies are at high‌ risk of aortic rupture‌​‌ and require the exclusion​​ of the diseased segment​​​‌ by the interposition of‌ a graft. TEVAR has‌​‌ a very high technical​​ success rate over 90%,​​​‌ and an acceptable rate‌ of post-procedural complications (4%‌​‌ mortality at 30 days).​​ Yet, the material properties​​​‌ of this graft have‌ been reported to be‌​‌ up to 10 times​​ less compliant than native​​​‌ vessels. This increased vascular‌ stiffness has been linked‌​‌ to increased cardiovascular morbidity​​ and mortality in the​​​‌ literature. In a context‌ of expansion of TEVAR‌​‌ indications to younger patients​​ with longer coverage of​​​‌ the aorta, this raises‌ the question of the‌​‌ morbidity on the long​​ term.

This study is​​​‌ the fruit of a‌ collaboration following the doctoral‌​‌ work of Alexandra Hauguel​​ (Patient-Specific Biomechanics of​​​‌ Thoracic Endovascular Aortic Repair‌ (TEVAR) under the supervision‌​‌ of A. Barakat and​​ S. Haulon, defended on​​​‌ 10/12/2024). The focus is‌ set on the potential‌​‌ impact of TEVAR on​​ the hemodynamics within the​​​‌ graft, but in particular‌ in the upstream ascending‌​‌ aorta (not covered by​​ the graft). The idea​​​‌ is to investigate if‌ a TEVAR could lead‌​‌ to negative feedback on​​ the heart with increased​​​‌ cardiac workload. The work‌ is conducted on a‌​‌ cohort of 13 patients​​ who all underwent a​​​‌ TEVAR procedure (7 of‌ them for a thoracic‌​‌ aortic aneurysm, 6 of​​ them for an aortic​​​‌ dissection). 4D flow MRI‌ data before the operation‌​‌ and 6 months after​​ have been acquired, which​​​‌ gives access to the‌ velocity field in the‌​‌ aorta over a cardiac​​ cycle. Tools have been​​​‌ developed to post-process the‌ obtained velocity fields, including‌​‌ required corrections of the​​ velocity field, and derived​​​‌ hemodynamic quantities such as‌ vorticity and helicity are‌​‌ investigated.

Collaborators: A. Hauguel,​​ X. Zhang, G. Cardillo,​​​‌ A. Barakat at LadHyX‌ - Ecole Polytechnique, A.‌​‌ Azarine at Hôpital Marie​​ Lannelongue

8.5.1 Multi-fidelity deep‌​‌ learning models to predict​​ Navier-Stokes solution in 2D​​​‌ domains

Participants: Francesco Songia‌, Irene Vignon-Clementel.‌​‌

Portal hypertension can be​​ reduced by inserting an​​​‌ artificial shunt to lower‌ the pressure. This procedure‌​‌ affects the hemodynamics of​​ the surrounding vessels. For​​​‌ a given patient, clinicians‌ don't know the exact‌​‌ size of the device​​ to control the pressure​​​‌ and flow distribution. We‌ want to provide a‌​‌ quick prediction through deep​​ learning methods of the​​​‌ pressure and velocity fields‌ to be able to‌​‌ evaluate a possible design​​ configuration.

Before facing the​​​‌ clinical application in 3D,‌ we have implemented a‌​‌ model that we applied​​ on a 2D synthetic​​​‌ dataset. The model is‌ based on a multi-fidelity‌​‌ pipeline to progressively learn​​ the stationary Navier-Stokes equations​​​‌ starting from reduced-order and‌ full Stokes approximation. We‌​‌ based our architectures on​​​‌ graph neural networks, and​ we combine them with​‌ transformers and space state​​ models. The learning process​​​‌ is guided by physical​ knowledge coming from the​‌ PDEs describing the system.​​ This results in improved​​​‌ accuracy and generalization properties​ across varying geometries 28​‌.

Collaborators: H. Talbot,​​ CentraleSupelec, Université Paris-Saclay, Inria​​​‌

8.5.2 Whole-body vascular transport​ and pharmacokinetics models: application​‌ to imaging

Participants: Jérôme​​ Kowalski, Dirk Drasdo​​​‌, Leo Donzil,​ Irene Vignon-Clementel.

A​‌ crucial aspect in the​​ surgical decision process is​​​‌ organ perfusion and functional​ assessment. In this context,​‌ a key area of​​ medical imaging is dynamic​​​‌ contrast-enhanced imaging. This typically​ involves a tracer (or​‌ contrast agent), which is​​ transported through the blood​​​‌ circulation with specific time​ dynamics as it passes​‌ through the various components​​ of the circulatory system.​​​‌ However, following a disease​ or a major surgery,​‌ the time dynamics of​​ the different elements are​​​‌ subject to change. A​ better understanding of the​‌ impact of a disease​​ on the time dynamics​​​‌ of a tracer would​ enhance the interpretation of​‌ the measured tracer signals,​​ thereby helping radiologists and​​​‌ surgeons to detect abnormal​ behaviors. In this project,​‌ we aim to understand​​ these dynamical signals through​​​‌ a combination of mathematical​ modeling of flow and​‌ transport, numerical simulations, determination​​ of model parameters from​​​‌ imaging data, and machine​ learning.

This year, we​‌ focused on data interpretation.​​ In particular, through our​​​‌ previous study on the​ validity constraints of classical​‌ reduced transport models, we​​ critically challenged the choice​​​‌ of model for interpreting​ dynamic contrast-enhanced images in​‌ the literature. Our study​​ highlights the possible presence​​​‌ of an apparent diffusion​ phenomenon that better reproduces​‌ the observed time dynamics​​ than molecular diffusion alone.​​​‌ The appearant diffusion phenomenon​ has been studied in​‌ synthetic trees and in​​ data from corrosion casts.​​​‌ Moreover, healthy and disease-altered​ vascular trees have been​‌ compared.

Finally, we have​​ enhanced the sensitivity analysis​​​‌ of the previously developed​ whole-body model of blood​‌ flow and transport by​​ utilizing surrogate modeling with​​​‌ an artificial neural network.​ This refined the parameter​‌ reduction choices and enabled​​ quantification of the identifiability​​​‌ of the chosen parameters​ through the profile likelihood​‌ analysis method, which leads​​ to the construction of​​​‌ confidence intervals around parameter​ estimates.

Collaborators: L. Sala​‌ (INRAE), J. R. van​​ der Vorst (LUMC, The​​​‌ Netherlands), C. Debbaut (UGent,​ Belgium)

8.5.3 Derivation and​‌ Validation of Compartment Models:​​ Implications for Dynamic Imaging​​​‌

Participants: Jérôme Kowalski,​ Dirk Drasdo, Irene​‌ Vignon-Clementel.

Compartment models​​ are mathematical representations of​​​‌ the transport of a​ chemical substance in the​‌ human body, assuming uniform​​ concentration within each compartment,​​​‌ which corresponds to a​ distinct body part. Widely​‌ used in pharmaceutical and​​ medical imaging through pharmacokinetic​​​‌ and tracer kinetic (TK)​ models, they are formulated​‌ as systems of time-dependent​​ ordinary differential equations. However,​​​‌ these models do not​ account for the spatial​‌ dependence of physiological processes​​ within individual compartments. This​​​‌ work aims toexplore the​ physical interpretation of these​‌ models through mathematical derivation​​ and to analyze their​​ limitations. Three TK models​​​‌ are derived from more‌ complex models regarding the‌​‌ physiological processes represented. The​​ derivation introduces the hypotheses​​​‌ relevant to these processes.‌ The most notable is‌​‌ the well-mixed hypothesis,resembling concentration​​ homogeneity in the region​​​‌ of interest. The hypotheses‌ are numerically tested by‌​‌ simulating the more complex​​ model and evaluating the​​​‌ reduced models' residuals. For‌ a given tracer molecule,‌​‌ the validity of TK​​ models is found to​​​‌ decrease as voxel size‌ increases.

This work 19‌​‌ proposes an algorithm to​​ compute the maximum voxel​​​‌ size, above which the‌ reduced model does not‌​‌ align with the physical​​ processes. Voxel sizes larger​​​‌ than 250 $\mu \mathrm{m‌}$ induce more than 3%‌​‌ misalignment error for highly​​ permeable vessels, while sizes​​​‌ smaller than 4 mm‌ induce less than 3%‌​‌ error in non-permeable vessels.​​ Tested on literature-derived datasets,​​​‌ these findings demonstrate that‌ molecular diffusion alone is‌​‌ not sufficient to model​​ a region of interest​​​‌ as a well-mixed compartment,‌ and strengthen the importance‌​‌ of high spatial resolution​​ imaging for improving TK​​​‌ parameters ground truth estimation.‌ Collaborator: L. Sala, INRAE‌​‌

8.5.4 In-silico modeling of​​ Hypoplastic Left Heart Syndrome​​​‌ patient as a clinical‌ tool for stage one‌​‌ planning surgery

Participants: Aseem​​ Milind Pradhan, Marie​​​‌ Haghebaert, Irene Vignon-Clementel‌.

The MEDITWIN project‌​‌ aims to develop a​​ preoperative surgical planning tool​​​‌ to support the medical‌ team in managing Hypoplastic‌​‌ Left Heart Syndrome (HLHS)​​ patients. HLHS is a​​​‌ complex congenital cardiac condition‌ which needs a series‌​‌ of surgeries to ensure​​ the survival of the​​​‌ patient. The first-stage palliation‌ (Norwood procedure) is especially‌​‌ critical, as it establishes​​ the physiological conditions necessary​​​‌ for both immediate survival‌ and eligibility for subsequent‌​‌ surgeries. Clinicians employ a​​ variety of surgical and​​​‌ interventional strategies informed by‌ individual patient data, institutional‌​‌ preference, and surgeon experience.​​ Currently morbidity and mortality​​​‌ remain substantial. Early reintervention‌ rates vary with 20-40‌​‌ percent of patients requiring​​ additional procedures before the​​​‌ stage-2 intervention. There is,‌ thus, an opportunity for‌​‌ improving outcomes with the​​ aid of virtual human​​​‌ twins.

The aim of‌ this work is to‌​‌ construct zero-dimensional (lumped parameter)​​ computational models representing the​​​‌ entire HLHS circulation before‌ and after the first-stage‌​‌ intervention. Further, the four​​ potential strategies for this​​​‌ stage will be evaluated,‌ including the traditional Norwood‌​‌ procedure with either a​​ Blalock-Thomas-Taussig or Sano shunt,​​​‌ Hybrid Norwood approach, and‌ fully catheter-based interventions as‌​‌ practiced at Hôpital Necker-Enfants​​ Malades, Paris. The model​​​‌ incorporates key cardiovascular compartments,‌ in particular the single-fiber‌​‌ model 16, allowing​​ assessment of how different​​​‌ procedural strategies affect relevant‌ biomarkers.

Collaborators: MEDITWIN project‌​‌ - Necker-Enfants Malades Hospital,​​ INRIA COMMEDIA team, 3DS​​​‌

8.6.1 How high-resolution‌​‌ agent-base models can improve​​ fundamental insights in tissue​​​‌ development and cell culturing‌ methods

Participants: Jiří Pešek‌​‌, Dirk Drasdo.​​

The fundamental understanding of​​​‌ how cells physically interact‌ with each other and‌​‌ their environment is key​​ to understanding their organisation​​​‌ in living tissues. Over‌ the past decades several‌​‌ computational methods have been​​​‌ developed to decipher emergent​ multi-cellular behaviors. In particular​‌ agent-based (or cell-based) models​​ that consider the individual​​​‌ cell as basic modeling​ unit tracked in space​‌ and time enjoy increasing​​ interest across scientific communities.​​​‌ We explored a particular​ class of cell-based models,​‌ so-called Deformable Cell Models​​ (DCMs), that allow to​​​‌ simulate the biophysics of​ the cell with high​‌ realism. Based on the​​ DCM we developped a​​​‌ model for organoid (cyst)​ growth and performed further​‌ model simulations for various​​ other in vitro systems​​​‌ to demonstrate how the​ DCM can be used​‌ to guide interpretation of​​ bioengineering experiments 27.​​​‌

Collaborators: Paul Van Liedekerke​ (University of Ghent), Kevin​‌ Alessandri (Treefrog Therapeutics)

8.6.2​​ Space Negotiation strategies of​​​‌ cells invading extracellular matrix:​ Lessons from a physics-based​‌ model

Participants: Dirk Drasdo​​.

Cell migration in​​​‌ Extracellular matrix (ECM) or​ ECM-like environments is driven​‌ by the pulling of​​ cells at ECM fibers.​​​‌ However, this same ECM​ also acts as a​‌ physical barrier for cell​​ migration if cells experience​​​‌ the ECM as mechanical​ obstacles at too high​‌ densities. To migrate, cells​​ need to maneuver through​​​‌ the ECM to find​ holes they can squeeze​‌ through their shape. Alternatively,​​ cells can mechanically adapt​​​‌ the elastic ECM and/or​ break down the ECM​‌ using matrix metalloproteinases (MMPs).​​ The combination of these​​​‌ mechanism results in cells​ that move optimal in​‌ medium dense environments and​​ which require MMP degradation​​​‌ and deformability to navigate​ highly dense environments. We​‌ studied within a novel​​ cell (agent)-based model if​​​‌ these three space negotiation​ mechanisms in combination with​‌ migration via pulling are​​ sufficient to explain larger​​​‌ scale observations in cell​ migration, namely single cell​‌ ECM invasion and multicellular​​ coordinated migration.

Collaborators: Andreas​​​‌ Buttenschön (University of Massachusetts​ Amherst), Margriet Palm (National​‌ Institute for Public Health​​ and the Environment, Netherlands​​​‌ (RIVM)), Philippe Chaviert (Institute​ Curie, Paris), Paul Van​‌ Liedekerke (University of Ghent)​​

8.7.1 Towards​ standardisation of center-based models​‌

Participants: Jieling Zhao,​​ Jiří Pešek, Jules​​​‌ Dichamp, Matteo Pedrazzi​, Dirk Drasdo.​‌

There is a wide​​ agreement for models and​​​‌ simulation tools at the​ level of intracellular models​‌ and at the level​​ of continuum descriptions (organ​​​‌ level, body level, flow,​ transport, biomechanical models, …).​‌ At the histological scale​​ between the intracellular and​​​‌ the organ level, tissues​ are modelled increasingly by​‌ agent-based models, that display​​ each individual cell. While​​​‌ models displaying cells on​ space-fixed lattices cannot capture​‌ the biomechanics correctly by​​ construction (Van Liedekerke et.​​​‌ al., 2015), center-based models​ (CBMs), in which forces​‌ between cells are modelled​​ as forces between cell​​​‌ centers in continuum space​ are increasingly used to​‌ simulate growth and re-organisation​​ processes in multicellular tissue​​​‌ (as for example in​ the cancer use-case in​‌ EDITH). Conceptually, CBMs mimic​​ cells similarly as active​​​‌ slightly deformable adhesive colloidal​ particles.

Several software tools​‌ are being used for​​ this purpose (e.g. Chaste,​​​‌ Physicell/Physiboss, Biodynamo, TiSim, CompuTiX).​ The problem is that​‌ there is no standard​​ on how to formulate​​ the CBMs even at​​​‌ a basic level as‌ for example for the‌​‌ Navier-Stokes Equations in fluid​​ mechanics that form an​​​‌ accepted basis for the‌ simulation of isotropic homogeneous‌​‌ fluids. As a consequence,​​ the different CBMs implemented​​​‌ in different software tools‌ would lead to different‌​‌ answers for the same​​ question.

In a collaboration​​​‌ with Arnau Montagud and‌ other partners representing the‌​‌ software tools enumerated above,​​ a set of basic​​​‌ simple model unit problems‌ and use cases were‌​‌ developed for which different​​ tools should give the​​​‌ same answers and interpretations,‌ and simulations were performed‌​‌ to compare and, as​​ far as possible, unify​​​‌ the simulation results. This‌ line of research has‌​‌ further be pursued and​​ refined ([NtianiakouEtAl2025]).​​​‌

Main collaborators: Arnau Montagud‌ (CSIC), Alfonso Valencia (Barcelona‌​‌ Supercomputing Center.), Paul Van​​ Liedekerke (University of Ghent)​​​‌

8.7.2 CompuTiX

Participants: Jiří‌ Pešek, Jules Dichamp‌​‌, Charles Boulitrop,​​ Peter Kottman, Matteo​​​‌ Pedrazzi, Dirk Drasdo‌.

Information on progress‌​‌ in 2025 is given​​ in section 7.

8.7.3​​​‌ TiSim

Participants: Jieling Zhao‌, Dirk Drasdo.‌​‌

Information on progress see​​ section 7.

9.1.1 Heartflow​​

Participants: Francesco Songia,​​​‌ Tobias Schnirer-Nedjar, Raoul‌ Sallé de Chou,‌​‌ Irene Vignon-Clementel [correspondant].​​

This project is in​​​‌ collaboration with Hugues Talbot‌ (CentraleSupelec & INRIA OPIS),‌​‌ Laurent Najmann (ESIEE/G Eiffel​​ University) and the company​​​‌ Heartflow. The goal is‌ to generate heart perfusion‌​‌ maps by machine learning.​​ See the PhD thesis​​​‌ of Raoul Sallé de‌ Chou for more information‌​‌ [Thesis Salle De​​ Chou, 23].​​​‌

10.1.1‌​‌ Associate Teams in the​​ framework of an Inria​​​‌ International Lab or in‌ the framework of an‌​‌ Inria International Program

Collaboration​​ with CHUM in Montreal,​​​‌ co-advising of PhD student‌ Mahdi Rezaei with Prof.‌​‌ Gilles Soulez, MD PhD:​​ visit of Mahdi Rezaei​​​‌ with the joint-programme MITACS.‌

10.2.1 Visits of international​​ scientists

• Prof. Eduard Rohan​​​‌
  • Status
    Professor
  • Institution of‌ origin
    West Bohemia University‌​‌
  • Country
    Czech Republic
  • Dates​​
    5 days (January-February 2025)​​​‌ + 2 days (December‌ 2025)
  • Context of the‌​‌ visit
    co-advising of PhD​​ of Peter Kottman, ERC​​​‌ MoDeLLiver
• Lazaros Papamanolis
  • Status‌
    PhD student
  • Institution of‌​‌ origin
    Stanford University
  • Country​​
    USA
  • Dates
    2.5 months​​​‌ (June 15 - August‌ 31, 2025)
  • Context of‌​‌ the visit
    Collaborative project​​ with PhD student P.​​​‌ Varsos, Chateaubriand STEM Fellowship‌

10.2.2 Visits to international‌​‌ teams

10.3.1 Horizon​​​‌ Europe

Participants: Friederike Schäfer‌, Irene Vignon-Clementel,‌​‌ Dirk Drasdo, Matteo​​ Pedrazzi, Jules Dichamp​​​‌, Francesco Songia,‌ Pavlos Varsos.

ARTEMIs‌​‌

ARTEMIS project on cordis.europa.eu​​

• Title:
  AcceleRating the Translation​​​‌ of virtual twins towards‌ a pErsonalised Management of‌​‌ fatty lIver patients
• Duration:​​​‌
  From January 1, 2024​ to December 31, 2027​‌
• Partners:
  • INSTITUT NATIONAL DE​​ RECHERCHE EN INFORMATIQUE ET​​​‌ AUTOMATIQUE (INRIA), France
  • MEDIZINISCHE​ UNIVERSITAET WIEN, Austria
  • IMPERIAL​‌ COLLEGE OF SCIENCE TECHNOLOGY​​ AND MEDICINE, United Kingdom​​​‌
  • FONDATION CARDIOMETABOLISME NUTRITION, France​
  • ASSISTANCE PUBLIQUE HOPITAUX DE​‌ PARIS, France
  • ALBERT-LUDWIGS-UNIVERSITAET FREIBURG​​ (UFR), Germany
  • DEUTSCHES KREBSFORSCHUNGSZENTRUM​​​‌ HEIDELBERG (GERMAN CANCER RESEARCH​ CENTER), Germany
  • CLINIQUES UNIVERSITAIRES​‌ SAINT-LUC ASBL, Belgium
  • BETTHERA​​ SRO (BETTHERA), Czechia
  • FUNDACION​​​‌ PARA LA INVESTIGACION DEL​ HOSPITAL UNIVERSITARIO LA FE​‌ DE LA COMUNIDAD VALENCIANA​​ (HULAFE), Spain
  • UNIVERSITAET LEIPZIG​​​‌ (ULEI), Germany
  • MEDICAL RESEARCH​ INFRASTRUCTURE DEVELOPMENT AND HEALTH​‌ SERVICES FUND BY THE​​ SHEBA MEDICAL CENTER (Sheba​​​‌ Research Fund), Israel
  • BOURNEMOUTH​ UNIVERSITY, United Kingdom
  • CHARITE​‌ - UNIVERSITAETSMEDIZIN BERLIN, Germany​​
  • UNIVERSITATSKLINIKUM HEIDELBERG (UKHD), Germany​​​‌
  • EUROPEAN LIVER PATIENTS ASSOCIATION​ (ELPA), Belgium
  • MEDEXPRIM (MEDEXPRIM),​‌ France
  • FUNDACIO HOSPITAL UNIVERSITARI​​ VALL D'HEBRON - INSTITUT​​​‌ DE RECERCA (VHIR), Spain​
  • UNIVERSITA DEGLI STUDI DI​‌ ROMA LA SAPIENZA (UNIROMA1),​​ Italy
  • MATICAL INNOVATION SL,​​​‌ Spain
  • UNIVERSITATSKLINIKUM JENA, Germany​

  Inria contact: Dirk Drasdo​‌

  Clinical Coordinator: Vlad Ratziu​​

  Clinical Data Coordinator: Raul​​​‌ Herance

  Scientific Coordinator: Irene​ Vignon-Clementel

  Summary: The ARTEMIs​‌ project aims to consolidate​​ existing computational mechanistic and​​​‌ machine-learning models at different​ scales to deliver ‘virtual​‌ twins’ embedded in a​​ clinical decision support system​​​‌ (CDSS). The CDSS will​ provide clinically meaningful information​‌ to clinicians, for a​​ more personalised management of​​​‌ the whole spectrum of​ Metabolic Associated Fatty Liver​‌ Disease (MAFLD). MAFLD, with​​ an estimated prevalence of​​​‌ about 25%, goes from​ an undetected sleeping disease,​‌ to inflammation (hepatitis), to​​ fibrosis development (cirrhosis) and/or​​​‌ hepatocellular carcinoma (HCC), decompensated​ cirrhosis and HCC being​‌ the final stages of​​ the disease. However, many​​​‌ MAFLD patients do not​ die from the liver​‌ disease itself, but from​​ cardiovascular comorbidities or complications.​​​‌

The ARTEMIs will contribute​ to the earlier management​‌ of MAFLD patients, by​​ prognosing the development of​​​‌ more advanced forms of​ the disease and cardiovascular​‌ comorbidities, promoting active surveillance​​ of patients at risk.​​​‌ The system will predict​ the impact of novel​‌ drug treatments or procedures,​​ or simply better life​​​‌ habits. The system will​ therefore not only serve​‌ as a clinical decision​​ aid tool, but also​​​‌ as an educational tool​ for patients, to promote​‌ better nutritional and lifestyle​​ behaviors.

In more advanced​​​‌ forms of the disease,​ therapeutic interventions include TIPPS​‌ to manage portal hypertension,​​ partial hepatectomy, partial or​​​‌ complete liver transplant. ARTEMIs​ will contribute to predict​‌ per- or post-intervention heart​​ failure, building on existing​​​‌ microcirculation hemodynamics models.

The​ model developers will benefit​‌ from a large distributed​​ patient cohort and data​​​‌ exploration environment to identify​ patterns in data, draw​‌ new theories on the​​ liver-heart metabolic axis and​​​‌ validate the performance of​ their models.

The project​‌ includes a proof-of-concept feasibility​​ study assessing the utility​​​‌ of the integrated virtual​ twins and CDSS in​‌ the clinical context.

10.3.2​​ H2020 projects

Participants: Ramdane​​​‌ Bessaid, Clemence Finotto​, Sylvain Freud,​‌ Peter Kottman, Jerome​​ Kowalski, Mahdi Rezaei​​​‌, Tobias Schnirer,​ Francesco Songia, Pavlos​‌ Varsos, Irene Vignon-Clementel​​.

MoDeLLiver

ERC MoDeLLiver​​

H2020 ERC consolidator grant​​​‌ MoDeLLiver is about 'Numerical‌ modelling of hemodynamics and‌​‌ pharmacokinetics for clinical translation'.​​ Surgical interventions are based​​​‌ on patient data, and‌ although they require careful‌​‌ planning, they may be​​ revised during surgery. To​​​‌ better predict surgery outcome,‌ several aspects must be‌​‌ considered, including the local​​ point of intervention, whole​​​‌ organ perfusion and function‌ as well as their‌​‌ interaction with the entire​​ circulation. To address this​​​‌ complexity, the EU-funded MoDeLLiver‌ project aims to develop‌​‌ a haemodynamic model to​​ guide surgical interventions in​​​‌ the lung and liver.‌ Researchers will also employ‌​‌ an injected substance model​​ to unravel the link​​​‌ between non-invasive medical imaging‌ and organ perfusion and‌​‌ function: this will be​​ very useful to parameterise​​​‌ the model prior to‌ the patient's intervention. The‌​‌ new modelling tool is​​ expected to bring personalised​​​‌ surgical simulation a step‌ closer to reality. 10/2020-09/2026‌​‌

Collaborators are the groups​​ of E. Vibert, N.​​​‌ Golse, S. Roullet (APHP-Hop.‌ P Brousse, France), E.‌​‌ Rohan (U of West​​ Bohemia, Czech Republic), G.​​​‌ Soulez (CHUM, Canada), C.‌ Debbaut (U. Ghent, Belgium),‌​‌ J.r. Van der Vorst​​ (LUMC, The Netherlands), W.​​​‌ Huberts & F. Van‌ de Vosse (TuE, The‌​‌ Netherlands), S. Hascoet &​​ O. Mercier (Marie Lannelongue​​​‌ Hosp, France).

10.4.1 Other‌​‌ european programs/initiatives

• Participants: Jieling​​ Zhao, Dirk Drasdo​​​‌ [Correspondant].

  BMBF-LiSyM-Cancer

  BMBF‌ “LiSyM-CANCER” (liver systems medicine‌​‌ of cancer). This project​​ followed the project LiSyM​​​‌ and establishes liver systems‌ medicine approaches to understand‌​‌ progression in chronic liver​​ disease towards Hepatocellular cancer.​​​‌ The project is a‌ large network project linking‌​‌ many partners all over​​ Germany.

  Collaborators include the​​​‌ groups of Steven Dooley‌ (University Hospital Mannheim, Germany),‌​‌ Jan Hengstler (Leibniz Institute​​ IFADO, Dortmund, Germany), Johannes​​​‌ Bode (University Hospital Düsseldorf,‌ Germany)

11.1.1​​ Scientific events: organisation

11.1.2‌ Scientific events: selection

11.1.3​​ Journal

11.1.4​‌ Invited to events

• I.​​ Vignon-Clementel, represented the EU​​​‌ project ARTEMIs, European Virtual​ Human Twins Initiative high-level​‌ event, Brussels, Belgium, Oct​​ 21rst 2025
• I. Vignon-Clementel,​​​‌ Invited Member, European Commission​ (DG-Connect): ApplyAI Strategy: Healthcare​‌ Delivery Expert Meeting, June​​ 5th 2025
• I. Vignon-Clementel,​​​‌ WIC (surgical innovation week-end),​ June 27th-29th 2025
• I.​‌ Vignon-Clementel, organized research workshop​​ on digital twins, IHU​​​‌ Sepsis days, Dec 11th​ 2025

11.1.5 Leadership within​‌ the scientific community

• INRAE​​ just published its report​​​‌ on digital twins 32​: Irene Vignon-Clementel represented​‌ Inria in this working​​ group.
• Dirk Drasdo is​​​‌ associated with IfADo Leibniz​ Institute directing co-workers within​‌ the German national-wide project​​ LiSyM-Cancer.

11.1.6 Scientific expertise​​​‌

• I. Vignon-Clementel, VPHi (virtual​ physiological human institute), Board​‌ of Directors (representing Inria),​​ Europe
• I. Vignon-Clementel, VPHi​​​‌ clinical working group member​ (monthly meetings), Europe
• I.​‌ Vignon-Clementel is member of​​ the Advisory Board, EPSRC​​​‌ Healthcare Technologies NetworkPlus–BIOREME project​ (UK)
• I. Vignon-Clementel is​‌ the academic presentative in​​ Initiative Biomed in-silico France​​​‌ taskforce
• I. Vignon-Clementel, Scientific​ Advisory Board (Inria member),​‌ PariSanté Campus, France

11.1.7​​ Research administration

• Irene Vignon-Clementel​​​‌ is scientific coordinator of​ the ARTEMIs (EU Horizon​‌ Europe) project, and is​​ coleader of the use​​​‌ case 3 within that​ grant
• Irene Vignon-Clementel, mentoring​‌ at Inria Saclay
• Dirk​​ Drasdo is Coordinator for​​​‌ WP6-Artemis (multiscale modeling) and​ Co-coordinator of a medical​‌ usecase.
• Dirk Drasdo is​​ in the leadership team​​​‌ of the grant LiSyM-Cancer-2​ (German ministry for research)​‌
• Dirk Drasdo is modeling​​ coordinator of the subproject​​​‌ CTIP-HCC in LiSyM-Cancer-2

11.2.1​​ Supervision

• CentraleSupélec Parcours Recherche​​​‌ project in progress: C.​ Finotto: "Improved personalization of​‌ a 0D hemodynamic model​​ for liver surgery outcome",​​​‌ Oct. 2023 - Sept.​ 2025, supervisors: Roel Meiburg,​‌ I. Vignon-Clementel
• Licence 3​​ Internship: J. Dersoir: "Vascular​​​‌ Network Generation", Jun. 2025,​ supervisors: Jules Dichamp, Peter​‌ Kottman, Dirk Drasdo
• Master​​ 1 Internship: L. Donzil:​​​‌ "Multi-scale Segmentation Methods for​ Liver Medical Imaging", Jun.​‌ 2025 - Aug. 2025,​​ supervisors: Jérôme Kowalski, Peter​​​‌ Kottman, Irene Vignon-Clementel
• PhD​ in progress: M. Pedrazzi,​‌ "A model of liver​​ disease progression", Nov. 2024​​​‌ - present, supervisors: D.​ Drasdo and Marta Cascante​‌ Serratosa (University of Barcelona)​​
• PhD in progress: S.-G.​​ Milani Malekzadeh, "Computational Modeling​​​‌ of Totally Percutaneous First-Stage‌ Palliation for Hypoplastic Left‌​‌ Heart Syndrome", Dec 2025​​ - present, supervisors: Abdul​​​‌ Barakat (Ecole Polytechnique), Irene‌ Vignon-Clementel
• PhD in progress:‌​‌ G. Martin, "Cerebral haemodynamic​​ response in children following​​​‌ post-ECMO carotid artery decannulation:‌ modelling of intracerebral blood‌​‌ flow", Sorbonne U., Nov​​ 2022 - present, supervisors:​​​‌ Sabine Vartan (APHP, Sorbonne‌ U.), Irene Vignon-Clementel
• PhD‌​‌ in progress: F. Songia,​​ "Reduced order modelling of​​​‌ hemodynamics for liver surgery‌ procedure", Oct. 2024 -‌​‌ present, supervisors: I. Vignon-Clementel,​​ N. Golse (AP-HP), H.​​​‌ Talbot (CentraleSupelec).
• PhD in‌ progress: P. Kottman, "Multilevel‌​‌ modeling of flow and​​ transport in liver lobules​​​‌ in health and disease",‌ Sep. 2023 - present,‌​‌ supervisors: I. Vignon-Clementel, D.​​ Drasdo & E. Rohan​​​‌ (UWB Pilsen, Czech Republic).‌
• PhD in progress: P.‌​‌ Varsos, "Multi-fidelity modelling of​​ vascular shunts and clinical​​​‌ applications", Jul. 2023 -‌ present, supervisors: I. Vignon-Clementel,‌​‌ S. Pant, N. Golse.​​
• PhD in progress: J.​​​‌ Kowalski, ”Whole-body vascular transport‌ and pharmaco-kinetics models: application‌​‌ to imaging, in particular​​ of liver ”, Dec.​​​‌ 2022 - present ,‌ supervisors: I. Vignon-Clementel, D.‌​‌ Drasdo & L. Sala​​ (INRAE).
• PhD in progress:​​​‌ R. Sallé de Chou,‌ ”Machine Learning based prediction‌​‌ of heart perfusion maps”,​​ Oct. 2021 - June​​​‌ 2025, supervisors: H. Talbot‌ (CentraleSupelec and Inria Opis‌​‌ team), I. Vignon-Clementel, L.​​ Najman (ESIEE Paris)
• PhD​​​‌ in progress: M. Rezaei‌ Adariani, ”Flow Dynamic Modelling‌​‌ to Assess the Accurate​​ Forces Scheme of Magnetic​​​‌ Drug Eluting Beads Navigated‌ by Magnetic Resonance Imaging”,‌​‌ Sep. 2021 - present,​​ supervisors: G. Soulez (CR-CHUM,​​​‌ Montreal, Canada), I. Vignon-Clementel‌

11.2.2 Juries

• I. Vignon-Clementel,‌​‌ Hiring committee for DR2​​ (senior permanent research position)​​​‌ for Inria, admission, 2025‌
• I. Vignon-Clementel, Hiring committee‌​‌ for Handicap CRCN (Junior​​ permanent research position) for​​​‌ Inria, Rocquencourt, France
• I.‌ Vignon-Clementel, HDR defense committee‌​‌ (member): Daniel Balvay, U.​​ Paris Cité, Paris, France,​​​‌ Nov 7th 2025
• I.‌ Vignon-Clementel, PhD defense committee‌​‌ (member): Pascalle Wijntjes, TU​​ Eindhoven, The Netherland, July​​​‌ 1rst 2025
• I. Vignon-Clementel,‌ PhD defense committee (co-advisor):‌​‌ Raoul Salle de Chou,​​ UPS, June 17th 2025​​​‌
• I. Vignon-Clementel, PhD defense‌ committee (external examinor): Sean‌​‌ Wong, Imperial College London,​​ Jan 14th 2025
• I.​​​‌ Vignon-Clementel, PhD CSI: Rodrigue‌ Domba, Villejuif, Oct 7th‌​‌ 2025
• I. Vignon-Clementel, PhD​​ CSI: Romain Lemore, Paris​​​‌ Sept 30th 2025
• I.‌ Vignon-Clementel, PhD CSI: Quentin‌​‌ Vanderbecq, Paris April 29th​​ 2025
• I. Vignon-Clementel, PhD​​​‌ CSI: Juan Varga Garcia,‌ Saclay June 13th 2025‌​‌
• I. Vignon-Clementel, PhD CSI:​​ Littisha Lawrance, U. Paris-Saclay,​​​‌ Sept 25th 2024
• I.‌ Vignon-Clementel, Research Training committee:‌​‌ Fergal Miskdjian, CentraleSupelec, June​​ 3rd 2025

11.2.3 Educational​​​‌ and pedagogical outreach

• F.‌ Songia co-organized a week‌​‌ of hands-on for high-school​​ students at Inria (June​​​‌ 2025)
• J. Kowalski, F.‌ Songia and M. Garreau:‌​‌ helped and prepared activities​​ at the Fête de​​​‌ la Science events (Science‌ days, 3-4th October 2025)‌​‌
• I. Vignon-Clementel, roundtable 'understanding​​ digital twins', Fête de​​​‌ la Science event (Science‌ days), PariSanté Campus, Oct‌​‌ 7th 2025

11.3.1 Specific official responsibilities​​​‌ in science outreach structures‌

I. Vignon-Clementel, scientific responsible‌​‌ of public outreach for​​​‌ the Inria Saclay IDF​ research center (until February​‌ 2025)

11.3.2 Productions (articles,​​ videos, podcasts, serious games,​​​‌ ...)

• I. Vignon-Clementel, Interview,​ Hospitalia Magazine, vol 69​‌ + online June 11th​​ 2025, France (>7K followers​​​‌ on LinkedIn)
• I. Vignon-Clementel​ and the communication services​‌ of Inria Saclay made​​ a Video for the​​​‌ Olympiade prizes on numerical​ sciences (high school students),​‌ Versailles-Montpellier-Toulouse Academy8, Jan 16th​​ 2025, France
• M. Garreau​​​‌ and I. Vignon-Clementel, BME​ seed grant video interview​‌ on YouTube

11.3.3 Conferences​​ and talks

• Amaury Facques​​​‌ & Irene Vignon-Clementel, webinaire​ Bernoulli Lab (Inria-APHP), March​‌ 5th 2025.
• Pavlos Varsos,​​ Oral presentation, European Society​​​‌ of Biomechanics Conference, Zurich,​ Switzerland, July 2025.
• Pavlos​‌ Varsos, Oral presentation, International​​ Conference on Computational Bioengineering,​​​‌ Rome, Italy, September 2025.​
• Pavlos Varsos, Poster presentation,​‌ Engineering 4 Health Annual​​ Forum, Palaiseau, France, November​​​‌ 2025.
• Peter Kottman, Poster​ presentation, 20th International Symposium​‌ on Computer Methods in​​ Biomechanics and Biomedical Engineering​​​‌ (CMBBE2025), Barcelona, Spain, September​ 2025.
• Peter Kottman, Poster​‌ presentation, Journées Math-Bio-Santé, Montpellier,​​ France, November 2025.
• Peter​​​‌ Kottman, Poster presentation, Engineering​ 4 Health Annual Forum,​‌ Palaiseau, France, November 2025.​​
• Morgane Garreau, Oral presentation,​​​‌ 20th International Symposium on​ Computer Methods in Biomechanics​‌ and Biomedical Engineering (CMBBE2025),​​ Barcelona, Spain, September 2025.​​​‌
• Morgane Garreau & Peter​ Kottman, Oral presentation, Research​‌ conference in continuum mechanics​​ course at AgroParisTech, Palaiseau,​​​‌ France, September 2025.
• Morgane​ Garreau, Oral presentation, Lab​‌ seminar at Laboratoire BMBI​​ (BioMécanique et BioIngénierie, UMR​​​‌ CNRS 7338) of Université​ Technique de Compiègne, Compiègne,​‌ France, September 2025.
• Morgane​​ Garreau, Oral presentation, 4èmes​​​‌ Journées annuelles du GDR​ MECABIO Santé, Avignon, France,​‌ November 2025.
• Morgane Garreau,​​ Poster presentation, Engineering 4​​​‌ Health Annual Forum, Palaiseau,​ France, November 2025.
• Francesco​‌ Songia, Poster presentation, Engineering​​ 4 Health Annual Forum,​​​‌ Palaiseau, France, November 2025.​
• Francesco Songia, Poster presentation,​‌ IP Paris PhD students​​ Welcome Day, Palaiseau, France,​​​‌ December 2025.
• Francesco Songia,​ Oral presentation, GDR MePhy​‌ & I-Gaia Day: Machine​​ Learning in mechanics and​​​‌ physics, Paris, France, December​ 2025.
• Matteo Pedrazzi, Poster​‌ presentation, IP Paris PhD​​ students Welcome Day, Palaiseau,​​​‌ France, December 2025.
• Aseem​ Milind Pradhan, Oral presentation,​‌ 2nd Inria MEDITWIN Scientific​​ Day, Strasbourg, France, September​​​‌ 2025.
• Friederike Schäfer, Oral​ presentation, 6th International Conference​‌ on Uncertainty Quantification in​​ Computational Science and Engineering​​​‌ (UNCECOMP), Rhodes, Greece, June​ 2025.
• Friederike Schäfer, Poster​‌ presentation, Sex Differences in​​ CVD Summit, London, United​​​‌ Kingdom, September 2025.
• Friederike​ Schäfer, Poster presentation, Engineering​‌ 4 Health Annual Forum,​​ Palaiseau, France, November 2025.​​​‌
• Tobias Schnirer, Poster presentation,​ 20th International Symposium on​‌ Computer Methods in Biomechanics​​ and Biomedical Engineering (CMBBE2025),​​​‌ Barcelona, Spain, September 2025.​
• Jérôme Kowalski, Oral presentation,​‌ 3rd Digital Twin in​​ Engineering Conference, Paris, France,​​​‌ February 2025.
• Irene Vignon-Clementel,​ invited talk, Health &​‌ Tech Summit, Marseilles, France,​​ Dec 16th-17th 2025
• Irene​​​‌ Vignon-Clementel, invited talk, IHU​ Sepsis days, Annecy, France,​‌ Dec 11th 2025
• Irene​​ Vignon-Clementel, invited presentation, 4èmes​​​‌ Journées annuelles du GDR​ MECABIO Santé, Avignon, France,​‌ November 2025.
• Irene Vignon-Clementel,​​ roundtable, E4H day, IP​​​‌ Paris, Nov 20th 2025,​ Palaiseau, France
• Irene Vignon-Clementel,​‌ presented the EU project​​ ARTEMIs with Laura Munoz​​ (MATICAL) at European Virtual​​​‌ Human Twins Initiative side-event,‌ Oct 2nd 2025, Brussels,‌​‌ Belgium
• Irene Vignon-Clementel, invited​​ talk, AI & Transplantation,​​​‌ Rencontres Agence de Biomédecine,‌ Oct 3rd 2025, Paris,‌​‌ France
• Irene Vignon-Clementel, invited​​ presentation, iSi Health, Sept​​​‌ 12th 2025, Leuven, Belgium.‌
• Irene Vignon-Clementel, invited participant,‌​‌ Google for Health symposium,​​ July 3rd 2025, Paris,​​​‌ France
• Irene Vignon-Clementel, Seminar,‌ Inria Europe, July 3rd‌​‌ 2025, Paris, France
• Irene​​ Vignon-Clementel, Keynote, Digital Twins​​​‌ in Oncology symposium -‌ PSCC Connect - June‌​‌ 24th 2025, Villejuif, France​​
• Irene Vignon-Clementel, presentation, workshop​​​‌ on digital twins in‌ health, French National Academy‌​‌ of Medicine, June 19th​​ 2025, Paris, France
• Irene​​​‌ Vignon-Clementel, Seminar, Computational Mechanics,‌ June 16th 2025, IP‌​‌ Paris, Palaiseau, France
• Irene​​ Vignon-Clementel, Seminar, FHU Predicare,​​​‌ June 12th 2025, online,‌ France
• Irene Vignon-Clementel, Keynote,‌​‌ InnovaHeart, March 26th 2025,​​ Paris, France
• Irene Vignon-Clementel,​​​‌ Seminar (biomechanics), Ecole Polytechnique,‌ Jan 6th 2025, Palaiseau,‌​‌ France

11.3.4 Others science​​ outreach relevant activities

• J.​​​‌ Kowalski, F. Songia and‌ M. Garreau: presented SIMBIOTX‌​‌ team, their research and​​ the daily life of​​​‌ a researcher to middle‌ school and high school‌​‌ students (observation internships), Inria​​ Saclay, January & June​​​‌ & December 2025.
• F.‌ Songia: CHICHE!, meeting and‌​‌ presentation to high-school students,​​ Mars 2025
• F. Songia:​​​‌ Rendez-vous des Jeunes Mathématiciennes‌ et Informaticiennes (RJMI, Young‌​‌ mathematicians and computer scientists​​ meeting). Young girls in​​​‌ high school ateliers &‌ supervision, Inria Saclay, Oct‌​‌ 20th-21st 2025.
• M. Garreau:​​ conference for high school​​​‌ students, Mois des mathématiques‌ appliquées et industrielles organized‌​‌ by Société de Mathématiques​​ Appliquées et Industrielles (SMAI),​​​‌ Lycée Hemingway, Nîmes, France,‌ November 2025.
• I. Vignon-Clementel:‌​‌ conference at Rendez-vous des​​ Jeunes Mathématiciennes et Informaticiennes​​​‌ (RJMI, Young mathematicians and‌ computer scientists meeting). Young‌​‌ girls in high school​​ ateliers & supervision, Inria​​​‌ Saclay, Oct 20th 2025.‌
• I. Vignon-Clementel: high-school conference‌​‌ and research exercise, Lycée​​ Dumas, Rueil Malmaison, France,​​​‌ Dec 5th 2025.

International journals

• 15 article‌​‌O.Omar Ali,​​ A.Alexandre Bône,​​​‌ C.Caterina Accardo,‌ B.Belkacem Acidi,‌​‌ A.Amaury Facque,​​ P.Paul Valleur,​​​‌ C.Chady Salloum,‌ M.-M.Marc-Michel Rohe,‌​‌ I.Irene Vignon-Clementel and​​ E.Eric Vibert.​​​‌ Automatic Generation of Liver‌ Virtual Models with Artificial‌​‌ Intelligence.Annals of​​ SurgeryApril 2025HAL​​​‌DOIback to text‌
• 16 articleM.Marie‌​‌ Haghebaert, P.Pavlos​​ Varsos, R.Roel​​​‌ Meiburg and I.Irene‌ Vignon-Clementel. A comparative‌​‌ study of lumped heart​​ models for personalized medicine​​​‌ through sensitivity and identifiability‌ analysis.The Journal‌​‌ of Physiology2025HAL​​DOIback to text​​​‌back to text
• 17‌ articleJ.John Hanna‌​‌, P.Pavlos Varsos​​, J.Jérôme Kowalski​​​‌, L.Lorenzo Sala‌, R.Roel Meiburg‌​‌ and I.Irene Vignon-Clementel​​. A comparative analysis​​​‌ of metamodels for 0D‌ cardiovascular models, and pipeline‌​‌ for sensitivity analysis, parameter​​​‌ estimation, and uncertainty quantification​.Computers in Biology​‌ and Medicine193July​​ 2025, 110381HAL​​​‌DOIback to text​back to text
• 18​‌ articleJ. M.John​​ M Hanna, P.​​​‌Pavlos Varsos, J.​Jérôme Kowalski, L.​‌Lorenzo Sala, R.​​Roel Meiburg and I.​​​‌Irene Vignon-Clementel. A​ comparative analysis of metamodels​‌ for lumped cardiovascular models,​​ and pipeline for sensitivity​​​‌ analysis, parameter estimation, and​ uncertainty quantification.Computers​‌ in Biology and Medicine​​1932025, 110381​​​‌HALDOI
• 19 article​J.Jérôme Kowalski,​‌ L.Lorenzo Sala,​​ D.Dirk Drasdo and​​​‌ I.Irene Vignon-Clementel.​ Derivation and validation of​‌ compartment models: implications for​​ dynamic imaging.IEEE​​​‌ Transactions on Medical Imaging​November 2025. In​‌ press. HALDOIback​​ to text
• 20 article​​​‌F.Frederik Link,​ Y.Yujia Li,​‌ J.Jieling Zhao,​​ S.Stefan Munker,​​​‌ W.Weiguo Fan,​ Z.Zeribe Nwosu,​‌ Y.Ye Yao,​​ S.Shanshan Wang,​​​‌ C.Chenjun Huang,​ R.Roman Liebe,​‌ S.Seddik Hammad,​​ H.Hui Liu,​​​‌ C.Chen Shao,​ C.Chunfang Gao,​‌ B.Bing Sun,​​ N.Natalie Török,​​​‌ H.Huiguo Ding,​ M. P.Matthias Pa​‌ Ebert, H.Honglei​​ Weng, P.Peter​​​‌ ten Dijke, D.​Dirk Drasdo, S.​‌Steven Dooley and S.​​Sai Wang. ECM1​​​‌ attenuates hepatic fibrosis by​ interfering with mediators of​‌ latent TGF-$1$ activation​​.Gut743​​​‌March 2025, 424-439​HALDOIback to​‌ text
• 21 articleW.​​Weiqiang Liu, M.​​​‌Morgane Garreau, L.​Lazaros Papamanolis, C.​‌Christian Kassasseya, K.-A.​​Kim-Anh Nguyen-Peyre, N.​​​‌Nour Belkeziz, S.​Suzanne Verlhac, P.​‌Pablo Bartolucci and I.​​Irene Vignon-Clementel. In-silico​​​‌ modelling of cerebral vasculopathy​ risk for children with​‌ sickle cell disease.​​Multidisciplinary Biomechanics Journal49th​​​‌ congress of the Société​ de BiomécaniqueJanuary 2025​‌HALDOIback to​​ text
• 22 articleI.​​​‌Irene Vignon-Clementel. Deep-learning​ and hemodynamics models: efficient​‌ complementarity for disease diagnosis​​ and surgery planning.​​​‌Multidisciplinary Biomechanics Journal49th​ congress of the Société​‌ de BiomécaniqueJanuary 2025​​HALDOIback to​​​‌ text

Doctoral dissertations and​ habilitation theses

• 23 thesis​‌R.Raoul Sallé de​​ Chou. A myocardial​​​‌ perfusion simulation pipeline based​ on machine-learning computational methods​‌.Université Paris-SaclayJune​​ 2025HALback to​​​‌ textback to text​

Reports & preprints

• 24​‌ miscS.Saskia Eckert​​, C.Christian Kassasseya​​​‌, M.Morgane Garreau​, C. J.Curie​‌ Jordane Kakpovi, I.​​Irène Vignon-Clementel, S.​​​‌Suzanne Verlhac, B.​Blanche Bapst, L.​‌Luca Scarcia, H.​​Henri Guillet, K.​​​‌Katy Dremont, S.​Smaïne Kouidri, F.​‌Frédéric Segonds, K.-A.​​ N.Kim-Anh Nguyen- Peyre​​​‌ and P.Pablo Bartolucci​. Additive Manufacturing of​‌ real-scale carotid artery models:​​ The content–container interaction in​​​‌ sickle cell disease-related cerebral​ vasculopathy.November 2025​‌HALDOIback to​​ text
• 25 miscM.​​Morgane Garreau, T.​​​‌Thomas Puiseux, R.‌Ramiro Moreno, S.‌​‌Solenn Toupin, D.​​Daniel Giese, F.​​​‌Franck Nicoud and S.‌Simon Mendez. Impact‌​‌ of partial echo on​​ 4D flow MRI: the​​​‌ insight from synthetic MRI‌.January 2026HAL‌​‌
• 26 miscM.Mathieu​​ de Langlard, T.​​​‌Tobias Schnirer, O.‌Olivier Trassard, A.‌​‌Antonietta Messina, N.​​Nassima Benzoubir, A.​​​‌Anne Dubart-Kupperschmitt, J.-C.‌Jean-Charles Duclos-Vallee, C.‌​‌Catherine Guettier, D.​​Dirk Drasdo and I.​​​‌Irene Vignon-Clementel. From‌ the Micro-to the Mesoscale‌​‌ 3D Reconstruction of the​​ Human Liver Architecture.​​​‌July 2025HALback‌ to text
• 27 misc‌​‌P.Paul van Liedekerke​​, J.Jiří Pešek​​​‌, K.Kevin Alessandri‌ and D.Dirk Drasdo‌​‌. How high-resolution agent-based​​ models can improve fundamental​​​‌ insights in tissue development‌ and cell culturing methods‌​‌.January 2026HAL​​back to text
• 28​​​‌ miscF.Francesco Songia‌, R.Raoul Sallé‌​‌ de Chou, H.​​Hugues Talbot and I.​​​‌Irene Vignon-Clementel. Multi-fidelity‌ graph-based neural networks architectures‌​‌ to learn Navier-Stokes solutions​​ on non-parametrized 2D domains​​​‌.December 2025HAL‌back to text