2025​​Activity reportProject-TeamARAMIS​​​‌

Benchmarking and​​​‌ improving deep generative models‌ for unsupervised anomaly detection‌​‌ in brain FDG PET​​

We have developed and​​​‌ validated deep generative modelling‌ approaches for unsupervised anomaly‌​‌ detection in brain FDG​​ PET, targeting early dementia​​​‌ biomarkers. To enable evaluation‌ without manual annotation, we‌​‌ introduced a simulation-based framework​​ for generating realistic anomalies​​​‌ and benchmarking pseudo-healthy image‌ reconstruction methods (Hassanaly et‌​‌ al., MELBA, 2024 23​​). We also conducted​​​‌ the first large-scale benchmarking‌ of 20 VAE variants‌​‌ for 3D PET anomaly​​ detection (Hassanaly et al.,​​​‌ JMI, 2025 24).‌ To enhance robustness, we‌​‌ leveraged population variability using​​ Z-score approaches (Solal et​​​‌ al., SPIE Medical Imaging,‌ 2024 125) and‌​‌ proposed model aggregation and​​ normalisation strategies (Solal et​​​‌ al., SPIE Medical Imaging,‌ 2026 96). In‌​‌ addition, we introduced the​​ first Bayesian flow network-based​​​‌ method for medical imaging,‌ which outperformed existing models‌​‌ and reduced false positives​​ in Alzheimer's disease detection​​​‌ (Roy et al., DGM@MICCAI,‌ 2025 95). Together,‌​‌ these contributions advance robust,​​ scalable biomarker detection for​​​‌ clinical neuroimaging.

Machine learning‌ for clinical MRI: quality‌​‌ control, harmonisation, and computer-aided​​ diagnosis

We have advanced​​​‌ a suite of machine‌ learning approaches tailored for‌​‌ the unique challenges of​​ routine clinical brain MRI​​​‌ available in large-scale clinical‌ data warehouses. For quality‌​‌ control, we developed convolutional​​ neural networks for the​​​‌ automatic quality control of‌ 3D T1-weighted MRI (Bottani‌​‌ et al., MedIA, 2022​​ 4). Building on​​​‌ this foundation, we implemented‌ a simulation-based transfer learning‌​‌ strategy, pre-training on research​​ data with simulated artefacts​​​‌ and fine-tuning on clinical‌ images, which substantially improved‌​‌ the detection of motion​​ artefacts (Loizillon et al.,​​​‌ MedIA, 2024 38)‌ and overall quality assessment‌​‌ (Loizillon et al, MELBA,​​ 2024 37) in​​​‌ 3D T1-weighted MRIs. Quality‌ control was extended to‌​‌ 3D FLAIR images through​​ a semi-supervised domain adaptation​​​‌ technique (Loizillon et al.,‌ MedIA, 2025 75).‌​‌ To address dataset heterogeneity,​​ we demonstrated that image-to-image​​​‌ translation could convert contrast-enhanced‌ to non-contrast-enhanced images, supporting‌​‌ the harmonisation of clinical​​ datasets and enabling reliable​​​‌ downstream analysis (Bottani et‌ al., BMC Medical Imaging,‌​‌ 2024 5). In​​ the domain of computer-aided​​​‌ diagnosis, we evaluated standard‌ MRI-based machine learning models‌​‌ for dementia diagnosis and​​ found that their performance​​​‌ drops markedly and can‌ be confounded by quality‌​‌ and contrast agent effects​​ when transitioning from curated​​​‌ research cohorts to complex‌ clinical data (Bottani et‌​‌ al., MedIA, 2023 3​​). Lastly, we assessed​​​‌ unsupervised anomaly detection for‌ identifying age-related white matter‌​‌ hyperintensities in routine FLAIR​​ MRI, showing promise and​​​‌ robustness to quality variation,‌ but also underscoring current‌​‌ limitations for clinical implementation​​ (Loizillon et al., MIDL,​​​‌ 2024 39). Collectively,‌ these studies chart both‌​‌ the significant progress and​​ the outstanding barriers in​​​‌ deploying robust automated neuroimaging‌ tools in routine clinical‌​‌ settings.

Deep learning-based segmentation​​ of Parkinson's disease-related brain​​​‌ structures and lymphomas

While‌ deep learning has enabled‌​‌ significant progress in automatic​​ segmentation of anatomical and​​​‌ pathological brain structures, key‌ challenges for clinical robustness‌​‌ remain. We developed a​​​‌ pooling loss function introducing​ soft topology constraints, which​‌ reduced topological errors and​​ improved segmentation accuracy when​​​‌ annotations are scarce (Fu​ et al, SPIE Medical​‌ Imaging, 2023 120).​​ We also proposed a​​​‌ frequency disentangled learning strategy,​ separating image components during​‌ training, that boosted performance​​ for Parkinsonian nuclei segmentation​​​‌ in some settings (Fu​ et al, SPIE Medical​‌ Imaging, 2024 119;​​ Fu et al, Journal​​​‌ of Medical Imaging, 2024​ 118). For primary​‌ central nervous system lymphoma,​​ our comprehensive comparison showed​​​‌ that specialised nnU-Net models​ outperform general foundation models,​‌ and we released a​​ validated open-source nnU-Net tool​​​‌ that demonstrated robust performance​ across several centres (Fu​‌ et al, SPIE Medical​​ Imaging, 2025 91;​​​‌ Fu et al, Radiology:​ Imaging Cancer, 2025 71​‌). These works highlight​​ the need for anatomical​​​‌ priors, novel training methods,​ and comprehensive validation for​‌ reliable neuroimaging segmentation.

Confidence​​ intervals for performance evaluation​​​‌ in medical image segmentation​

In medical image segmentation,​‌ the lack of systematic​​ reporting on the precision​​​‌ of performance metrics, especially​ confidence intervals, undermines the​‌ reliability and clinical translation​​ of artificial intelligence models.​​​‌ We have shown that​ small test sets, often​‌ used in the field,​​ result in very wide​​​‌ confidence intervals for metrics​ such as the Dice​‌ coefficient, which limits interpretability​​ and comparability (El Jurdi​​​‌ and Colliot, IEEE ISBI,​ 2023 122). Our​‌ empirical and simulation studies​​ in brain MRI segmentation​​​‌ indicate that parametric and​ bootstrap methods produce similar​‌ confidence intervals, and that​​ segmentation typically requires fewer​​​‌ cases than classification to​ reach a given precision​‌ (El Jurdi et al.,​​ MedIA, 2025 70).​​​‌ Analysing major conferences with​ our international collaborators, we​‌ found that confidence intervals​​ are seldom reported, and​​​‌ differences between top-performing models​ are often not statistically​‌ significant when these intervals​​ are considered (Christodoulou et​​​‌ al., MICCAI, 2024 9​). We further identified​‌ that for median estimates,​​ some bootstrap approaches provide​​​‌ unreliable coverage, highlighting the​ need for careful methodological​‌ choices (André et al.,​​ BRIDGE MICCAI Workshop, 2025​​​‌ 89). Overall, our​ work strongly supports transparent​‌ and systematic confidence interval​​ reporting to improve the​​​‌ evaluation and clinical translation​ of segmentation models.

Disease course​​​‌ mapping

We designed a​ new class of mixed-effects​‌ models to learn distributions​​ of trajectories from a​​​‌ longitudinal dataset. The successive​ observations of a subject​‌ are seen as successive​​ points along a curve​​​‌ drawn on a multivariate​ Riemannian manifold. The subjects'​‌ curves all derive from​​ a common geodesics on​​​‌ the manifold, which summarises​ the progression scenario in​‌ the population. The derivation​​ comprises a time-reparameterisation of​​​‌ the geodesics to account​ for variations in the​‌ dynamics of changes and​​ exp-parallelisation to account for​​​‌ the distribution of the​ data at a given​‌ stage of progression. The​​ theoretical and computational foundations​​​‌ of these approaches were​ set up in the​‌ thesis and articles of​​ J.-B. Schiratti (Schiratti et​​​‌ al, JMLR, 2017 54​). During the PhD​‌ thesis of J. Ortholand,​​ the model was extended​​ to consider events in​​​‌ addition to longitudinal data.‌ This was first done‌​‌ for a model with​​ one feature and then​​​‌ the model was extended‌ to consider several features.‌​‌ In both cases, the​​ model was validated on​​​‌ simulated data with good‌ performance in comparison with‌​‌ other existing models. Modelling​​ the heterogeneous progression of​​​‌ chronic diseases requires methods‌ that extend beyond standard‌​‌ mixed-effects assumptions. We proposed​​ a probabilistic mixture extension​​​‌ of the Disease Course‌ Mapping model to capture‌​‌ distinct disease progression subtypes​​ within a population. This​​​‌ framework provided a robust‌ and interpretable approach for‌​‌ clustering patients according to​​ spatiotemporal disease dynamics, offering​​​‌ a valuable tool for‌ potential insights into precision‌​‌ medicine (Kaisaridi et al,​​ ISCB, 2025 104).​​​‌

Disease progression modelling for‌ trial design

Disease progression‌​‌ models can effectively model​​ the progression of diseases​​​‌ using a great variety‌ of multimodal longitudinal data‌​‌ and provide valuable insights​​ into the disease's clinical​​​‌ manifestations and progression. The‌ estimations obtained thanks to‌​‌ these models can inform​​ clinical trial design and​​​‌ facilitate more accurate prognosis‌ and individualised treatment strategies.‌​‌ In particular, we evaluated​​ AD Course Map, a​​​‌ statistical model predicting the‌ progression of neuropsychological assessments‌​‌ and imaging biomarkers for​​ a patient from current​​​‌ medical and radiological data‌ at early disease stages‌​‌ of Alzheimer's disease. We​​ showed that enriching the​​​‌ population with the predicted‌ progressors decreased the required‌​‌ sample size by 38%​​ to 50%, depending on​​​‌ trial duration, outcome, and‌ targeted disease stage, from‌​‌ asymptomatic individuals at risk​​ of AD to subjects​​​‌ with early and mild‌ AD (Maheux et al,‌​‌ Nature Communications, 2023 40​​). We also explored​​​‌ prediction-powered inference (PPI) and‌ its subsequent development, PPI++,‌​‌ which provide a novel​​ approach to standard statistical​​​‌ estimation by leveraging machine‌ learning systems to enhance‌​‌ unlabelled data with predictions.​​ We use this paradigm​​​‌ in clinical trials, where‌ the predictions are provided‌​‌ by disease progression models​​ such as those produced​​​‌ by Leaspy, providing prognostic‌ scores for all the‌​‌ participants as a function​​ of baseline covariates. The​​​‌ proposed method would empower‌ clinical trials by providing‌​‌ untreated digital twins of​​ the treated patients while​​​‌ remaining statistically valid (Poulet‌ et al, BMC Medical‌​‌ Research Methodology, 2025 81​​).

Methods for real-world​​​‌ health data

Our team‌ has developed rigorous methodological‌​‌ frameworks for leveraging large-scale​​ administrative health databases in​​​‌ neuroepidemiology research. We established‌ algorithms for disease identification‌​‌ and created novel approaches​​ to quantify diagnostic and​​​‌ encoding delays, providing empirical‌ tools to address temporal‌​‌ biases inherent in real-world​​ data. Through transnational collaborations​​​‌ across multiple healthcare systems‌ (France, Sweden, UK, Australia),‌​‌ we standardised case identification​​ methods and developed comparative​​​‌ analytical frameworks to distinguish‌ genuine epidemiological trends from‌​‌ database artefacts (Wei et​​ al, EBioMedicine, 2025 86​​​‌). Our work on‌ prodromal phases combined large-scale‌​‌ case-control designs with temporal​​ trajectory analysis to identify​​​‌ early disease markers while‌ controlling for multiple comparisons‌​‌ (Guinebretiere, thesis 108).​​ We applied target trial​​​‌ emulation methodology to draw‌ causal inferences from observational‌​‌ data, exploiting natural experiments​​​‌ like policy changes as​ sources of quasi-random treatment​‌ variation. We also analysed​​ the care pathway of​​​‌ patients from several diseases​ (Dibling et al, Neuroepidemiology,​‌ 2024 19). We​​ finally contributed to open-source​​​‌ infrastructure development in R​ (sndsTools) to standardise data​‌ processing and promote reproducibility.​​

Applications in chronic neurological​​​‌ diseases

The methodological advances​ were applied to a​‌ range of chronic neurological​​ diseases, including amyotrophic lateral​​​‌ sclerosis, Parkinson's disease, CADASIL,​ and cerebellar ataxia. These​‌ studies revealed key factors​​ influencing disease progression, such​​​‌ as sex, onset site,​ genetic modifiers, and comorbidities,​‌ and uncovered heterogeneity in​​ the trajectories of clinical​​​‌ symptoms and decline. The​ models provided new insights​‌ into early and distinct​​ patterns of progression, paving​​​‌ the way for better​ patient stratification and precision​‌ medicine in these disorders​​ 17, 74,​​​‌ 43.

Virtual staining and generative​​ modelling

A central scientific​​​‌ contribution is the invention​ of algorithms and methods​‌ for generating multiple virtual​​ immunohistochemical (IHC, antibody-reaction-based) stain​​​‌ images from a single​ H&E whole-slide image (mainly​‌ structural), using paired and​​ unpaired generative approaches. This​​​‌ patent formalises a scalable,​ trustworthy GenAI framework for​‌ virtual multi-staining in computational​​ pathology, enabling accurate prediction​​​‌ of immunostains (CD3, CD8,​ GIEMSA, CD163, AE1AE3, CD117,​‌ D2-40, CD15) directly from​​ routine H&E slides. The​​​‌ underlying model integrates explainable​ diffusion and specialised CycleGAN-based​‌ architectures constrained by biophysical​​ priors to ensure plausibility​​​‌ and interpretability. The foundational​ publication (Ounissi et al,​‌ PLoS Computational Biology, 2025​​ 46) underpinning these​​​‌ technologies demonstrates the scalability​ and reliability of our​‌ methods across multiple tissue​​ types and staining protocols.​​​‌

High-content microscopy and neurodegenerative​ pathology

Quantifying phagocytosis in​‌ dynamic, unstained cells is​​ crucial for studying neurodegenerative​​​‌ diseases, yet extremely challenging​ due to rapid cellular​‌ interactions, low contrast, and​​ acquisition artefacts in phase-contrast​​​‌ time-lapse microscopy. We introduced​ a scalable, real-time, end-to-end​‌ framework combining data quality​​ control, robust segmentation, and​​​‌ two complementary explainability modules​ that reveal deep learning​‌ decisions through visual attribution​​ and model simplification. This​​​‌ demonstrates that interpretability can​ enhance, rather than hinder,​‌ performance, yielding a more​​ efficient architecture and optimised​​​‌ execution. Applied to microglial​ phagocytosis in frontotemporal dementia​‌ (FTD), the method reveals​​ statistically significant alterations (FTD​​​‌ mutant cells being larger​ and more aggressive than​‌ controls) and achieves state-of-the-art​​ results across public benchmarks.​​​‌ To accelerate translational research,​ we have released the​‌ full pipeline and a​​ unique phagocytosis dataset, providing​​​‌ a reproducible foundation for​ future interpretable AI developments​‌ in neurodegenerative disease characterisation.​​ The methods and results​​​‌ were published in Ounissi​ et al, Scientific Reports,​‌ 2024 45, together​​ with the computational foundation​​​‌ for PhagoStat (github.com/ounissimehdi/PhagoStat​), a framework enabling​‌ interpretable quantification of cell​​ phagocytosis and neuroinflammation dynamics.​​​‌

Tauopathy analysis for refined​ Alzheimer's disease patient stratification​‌

Quantifying the distribution and​​ morphology of tau protein​​​‌ structures in brain tissue​ is essential for diagnosing​‌ Alzheimer's disease (AD) and​​ its variants and for​​​‌ refining patient stratification. We​ introduce a deep learning​‌ framework for semantic segmentation​​ of tau lesions, particularly​​ neuritic plaques, in WSI​​​‌ from post-mortem data provided‌ by the NEURO-CEB AP-HP‌​‌ / ICM human brain​​ tissue repository. We released​​​‌ ADNP-15, an open-source whole-slide‌ image dataset for neuritic‌​‌ plaque segmentation and stain​​ normalisation (Zhao et al,​​​‌ IRBM, 2025 62),‌ after a preliminary dataset‌​‌ and methodological publication (Jiménez​​ et al, MICCAI, 2022​​​‌ 121). Complementary work‌ in Ingrassia et al,‌​‌ J. Neuropathol. Exp. Neurol.,​​ 2024 26 has established​​​‌ new quantitative methods for‌ the interpretable segmentation of‌​‌ pathological structures such as​​ tau tangles and neuritic​​​‌ plaques using frugal, explainable‌ architectures.

Physics-informed modelling and‌​‌ multi-scale integration

We contributed​​ to the development of​​​‌ mechanistic models of tumour‌ oxygenation and hypoxia based‌​‌ on coupled PDEs constrained​​ by spatial imaging data​​​‌ (Kumar et al, Physics‌ Medicine and Biology, 2024‌​‌ 35). This work​​ exemplifies our broader approach​​​‌ of integrating data-driven deep‌ learning with mesoscopic physical‌​‌ modelling to enhance biological​​ realism, interpretability, and generalisation​​​‌ of predictive models. Such‌ physics-guided generative frameworks are‌​‌ being extended to longitudinal​​ tumour modelling and multi-scale​​​‌ image registration, linking MRI,‌ histology, and spatial transcriptomics‌​‌ in a consistent computational​​ space.

Responsible and frugal​​​‌ AI frameworks

The foundation‌ of this line of‌​‌ research is synthesised in​​ Racoceanu et al, Techniques​​​‌ de l'Ingénieur, 2022 50‌, which formalises the‌​‌ concept of Responsible AI​​ as applied to computational​​​‌ pathology, integrating explicability, traceability,‌ and human oversight, and‌​‌ demonstrates that explainability and​​ performance can coexist through​​​‌ frugal architectures (Ounissi et‌ al, Scientific Reports, 2024‌​‌ 45). This framework​​ underpins ongoing developments in​​​‌ multi-virtual staining (Ounissi et‌ al, PLoS Computational Biology,‌​‌ 2025 46), as​​ well as agentic AI​​​‌ architectures for multi-modal integration‌ (Kumar et al, Physics‌​‌ in Medicine and Biology​​ 2024 35).

Our​​​‌ results demonstrate a consistent‌ research trajectory, from theoretical‌​‌ foundations of responsible generative​​ AI to patented applications​​​‌ in virtual staining and‌ multi-scale biomedical data modelling,‌​‌ thereby solidifying our contributions​​ at the intersection of​​​‌ computational pathology, physics-informed AI,‌ and clinical translation.

High-dimensional statistics for​​​‌ brain imaging

Our group's‌ contributions in high-dimensional neuroimaging‌​‌ statistics are highlighted by​​ three major articles, each​​​‌ addressing foundational challenges with‌ innovative solutions and far-reaching‌​‌ implications for research and​​ clinical practice. In a​​​‌ first-of-its-kind analysis, we systematically‌ quantified how different cortical‌​‌ atlases capture trait-related brain​​ variance, revealing that atlas​​​‌ selection profoundly affects the‌ proportion of phenotypic variance‌​‌ explained by brain structure,​​ i.e. the morphometricity (Fürtjes​​​‌ et al, Cortex, 2023‌ 21). This work‌​‌ provides a rigorous, data-driven​​ framework for choosing atlases,​​​‌ directly impacting the reproducibility‌ and biological interpretability of‌​‌ neuroimaging findings across studies.​​ The second study pioneers​​​‌ a comprehensive benchmarking of‌ MRI processing pipelines, demonstrating‌​‌ for the first time​​ how pipeline choice influences​​​‌ morphometricity, replicability, and predictive‌ power (Delzant et al,‌​‌ Human Brain Mapping, 2025​​ 68). By identifying​​​‌ volume-based pipelines (e.g., FSLVBM)‌ as optimal for robustness‌​‌ and surface-based pipelines as​​ sources of unique but​​​‌ less consistent signals, it‌ sets a new standard‌​‌ for pipeline selection in​​​‌ high-dimensional neuroimaging, ensuring more​ reliable and generalisable results.​‌ The third article translates​​ these methodological advances into​​​‌ clinical impact by synthesising​ grey matter biomarkers for​‌ Alzheimer's disease progression and​​ cognitive decline (Couvy-Duchesne et​​​‌ al, Human Brain Mapping,​ 2024 13). Its​‌ meta-analytic approach not only​​ identifies robust, early-detection biomarkers​​​‌ but also bridges the​ gap between high-dimensional statistics​‌ and real-world applications, offering​​ actionable insights for early​​​‌ intervention and personalised medicine.​ Collectively, these studies redefine​‌ best practices in neuroimaging​​ analysis, from atlas and​​​‌ pipeline selection to clinical​ translation. Their originality lies​‌ in providing empirical, scalable​​ solutions to longstanding challenges​​​‌ in high-dimensional data, while​ their significance is amplified​‌ by their direct impact​​ on reproducibility, biological insight,​​​‌ and the potential for​ early disease detection, particularly​‌ in neurodegenerative disorders like​​ Alzheimer's.

Combining deep learning​​​‌ and advanced statistics to​ unveil the genetic underpinnings​‌ of imaging and clinical​​ phenotypes

Combining deep learning​​​‌ for imaging and advanced​ statistics for omics data,​‌ we were able to​​ unveil the genetic bases​​​‌ of various phenotypes that​ are highly relevant neuroscientifically​‌ and/or clinically but had​​ never been studied so​​​‌ far, due to lack​ of adequate tools. We​‌ built a deep learning​​ tool to measure choroid​​​‌ plexuses (Yazdan-Panah et al,​ NeuroImage: Clinical 61),​‌ a structure which is​​ highly relevant to multiple​​​‌ sclerosis and neuroinflammation, and​ used this tool to​‌ perform the first corresponding​​ genome-wide association (GWAS) that​​​‌ unveiled their complex genetic​ architecture (Yazdan-Panah, PhD thesis,​‌ 2025 109; Yazdan-Panah​​ et al, In preparation).​​​‌ In a similar spirit,​ we built a tool​‌ for automatic rating of​​ incomplete hippocampal inversion and​​​‌ extensively validated it across​ multiple cohorts and over​‌ 1,000 participants (Hemforth et​​ al, MELBA, 2024 25​​​‌). Its application allowed​ us to perform a​‌ GWAS which robustly identified​​ various genetic variants associated​​​‌ with this phenotype (Hemforth​ et al, under revision​‌ at Imaging Neuroscience). We​​ studied other intriguing anatomical​​​‌ variations of the basal​ temporal lobe, in particular​‌ topological sulcal connections. We​​ demonstrated their heritability and​​​‌ unveiled a sexual dismorphism​ as well as association​‌ with incomplete hippocampal inversions​​ (De Matos et al,​​​‌ Brain Structure and Function,​ 2023 42).

Learning​‌ multimodal representations of imaging​​ and transcriptomic data

We​​​‌ introduced a deep learning​ framework learn joint representations​‌ that can integrate multimodal​​ data from neuroimaging and​​​‌ transcriptomics. To that purpose,​ we proposed a multimodal​‌ variational autoencoder formulation. An​​ original feature of this​​​‌ approach is the introduction​ of a weak supervision​‌ to constraint the latent​​ space structure to reflect​​​‌ disease severity. The approach​ was applied to study​‌ the progression of genetic​​ forms of frontotemporal dementia​​​‌ and amyotrophic lateral sclerosis.​ It allowed building disease​‌ progression scores that were​​ validated against disease stage​​​‌ (Kmetzsch et al, IEEE​ J. Biomed. Health. Inf.,​‌ 2022 31).

Parkinson's​​​‌ disease and related disorders‌

We strengthened our activities‌​‌ on Parkinson's disease through​​ several complementary approaches. In​​​‌ collaboration with genetics specialists‌ at ICM and internationally,‌​‌ we identified genetic determinants​​ of cognitive decline (Faouzi​​​‌ et al, npj Parkinson's‌ disease, 2024 20).‌​‌ During Lydia Chougar's Inria-funded​​ secondment (neuroradiologist), we identified​​​‌ imaging biomarkers to differentiate‌ Parkinsonian syndromes (Chougar et‌​‌ al, Park & Rel​​ Dis, 2023 8)​​​‌ and demonstrated that MRI‌ markers significantly improve diagnostic‌​‌ performance compared to standard​​ clinical diagnosis (Chougar et​​​‌ al, Movement Disorders, 2024‌ 7).

Amyotrophic lateral sclerosis

Using‌​‌ a multivariable disease modelling​​ approach, we showed that​​​‌ sex and onset site‌ are important drivers of‌​‌ the progression of motor​​ function, BMI, and FVC​​​‌ decline in ALS patients‌ (Ortholand et al, ENCALS,‌​‌ 2022 123).

Parkinson's​​ disease progression and heterogeneity​​​‌

We reconstructed the temporal‌ cascade of Parkinson's disease,‌​‌ finding that the first​​ changes occurred in the​​​‌ contralateral putamen 13 years‌ before diagnosis, followed by‌​‌ changes in motor symptoms,​​ dysautonomia, and sleep (all​​​‌ before diagnosis), and finally‌ cognitive decline at diagnosis.‌​‌ The model revealed earlier​​ disease onset, earlier non-motor​​​‌ and later motor symptoms,‌ and more rapid cognitive‌​‌ decline in PD patients​​ with REM sleep behavioural​​​‌ disorder compared to those‌ without. Understanding this heterogeneity‌​‌ is key to deciphering​​ underlying pathophysiology and selecting​​​‌ homogeneous subgroups for precision‌ medicine (Di Folco et‌​‌ al, Movement Disorders, 2023​​ 17).

CADASIL

In​​​‌ cerebral autosomal dominant arteriopathy‌ with subcortical infarcts and‌​‌ leukoencephalopathy (CADASIL), the most​​ frequent genetic small artery​​​‌ brain disease, we demonstrated‌ gradual and heterogeneous decline‌​‌ in different clinical and​​ cognitive performances over patients'​​​‌ lifetimes. Two progression profiles‌ emerged: one rapid and‌​‌ early, the other delayed​​​‌ and slower. Male gender,​ low educational level, pathogenic​‌ variant location in EGFr​​ 1 to 6 domains,​​​‌ smoking, and arterial hypertension​ were identified as factors​‌ affecting clinical progression (Kaisaridi​​ et al, Neurology, 2025​​​‌ 74).

Ataxias

Studying​ the scale for the​‌ assessment and rating of​​ ataxia (SARA), the reference​​​‌ clinical scale for cerebellar​ ataxia, we found that​‌ seven of eight items​​ showed non-linear progression while​​​‌ the total score progressed​ linearly. Progression speed differed​‌ between most items, providing​​ crucial information for clinical​​​‌ trial design in upcoming​ therapeutic trials (Moulaire et​‌ al, Movement Disorders, 2023​​ 43).

Neuro-oncology: primary brain​ lymphomas

During Lucia Niccheli's​‌ Inria-funded secondment (neuroradiologist), we​​ launched a new research​​​‌ programme on primary central​ nervous system lymphoma, a​‌ rare, highly malignant brain​​ tumour. We developed an​​​‌ open source model for​ automatic segmentation of such​‌ tumours with extensive internal​​ and external validation (Fu​​​‌ et al, Radiology: Imaging​ Cancer, 2025 71).​‌ Current work builds on​​ this foundation to develop​​​‌ approaches for predicting treatment​ response.

Neuro-ophthalmology AI-based stroke​‌ detection

Through collaborative PhD​​ supervision of Ayse Gungor​​​‌ and Ilias Sarbout, and​ leveraging neuro-ophthalmology expertise at​‌ the Adolphe de Rothschild​​ Foundation Hospital, we developed​​​‌ deep learning models for​ rapid detection of hyperacute​‌ central retinal artery occlusion​​ from retinal fundus photographs​​​‌ (J. American Heart Association,​ 2025 22). Our​‌ results suggest that AI-based​​ analysis of retinal photographs​​​‌ could support emergency stroke​ pathways, facilitate timely fibrinolysis​‌ decisions, and improve secondary​​ stroke prevention pending further​​​‌ validation.

7.1.1 Clinica

• Name:​​
  Clinica
• Keywords:
  Neuroimaging, Brain​​​‌ MRI, MRI, Clinical analysis,‌ Image analysis, Machine learning‌​‌
• Functional Description:

  Clinica is​​ a software platform for​​​‌ clinical neuroscience research, enabling‌ multimodal brain image analysis‌​‌ of large-scale datasets. It​​ facilitates the application of​​​‌ advanced analysis pipelines to‌ diverse clinical studies. To‌​‌ this end, it integrates​​ a comprehensive set of​​​‌ processing tools for the‌ main neuroimaging modalities: anatomical‌​‌ MRI, diffusion MRI, and​​ PET.

  For each modality,​​​‌ Clinica enables the extraction‌ of various types of‌​‌ features (regional measures, parametric​​ maps, surfaces,...). These features​​​‌ can then be used‌ as input for statistical‌​‌ modeling or machine learning​​ methods. Processing pipelines are​​​‌ based on combinations of‌ open-source tools developed by‌​‌ the community. Clinica follows​​ the BIDS specification for​​​‌ input data and proposes‌ one for the storage‌​‌ of processed outputs. Clinica​​ is written in Python​​​‌ and leverages the Nipype‌ system for pipelining. It‌​‌ combines widely-used software for​​ neuroimaging data analysis (SPM,​​​‌ Freesurfer, ANTs, FSL, MRtrix...),‌ and machine learning (scikit-learn).‌​‌

• URL:
  http://­www.­clinica.­run
• Publications:
  hal-02308126​​, hal-04278898, hal-03728243​​​‌, hal-03513920, hal-02549242‌, hal-02132147, hal-02562504‌​‌, hal-01518785, hal-01578479​​, hal-01858384, hal-01907482​​​‌, hal-01654000, hal-02566361‌, hal-01950933
• Contact:
  Ninon‌​‌ Burgos
• Participants:
  Ninon Burgos,​​ Olivier Colliot, Alice Joubert,​​​‌ Adam Ismaili, Matthieu Joulot,‌ Maelys Solal, Hugues Roy,‌​‌ Michael Bacci, Simona Bottani,​​ Mauricio Diaz, Stanley Durrleman,​​​‌ Sabrina Fontanella, Nicolas Gensollen,‌ Pietro Gori, Jeremy Guillon,‌​‌ Ravi Hassanaly, Thomas Jacquemont,​​ Sophie Loizillon, Pascal Lu,​​​‌ Arnaud Marcoux, Tristan Moreau,‌ Alexandre Routier, Omar El‌​‌ Rifai, Jorge Samper Gonzalez,​​ Elina Thibeau-Sutre, Ghislain Vaillant,​​​‌ Junhao Wen
• Partners:
  Institut‌ du Cerveau et de‌​‌ la Moelle épinière (ICM),​​ CNRS, INSERM, Sorbonne Université​​​‌

7.1.2 ClinicaDL

• Keywords:
  Deep‌ learning, Neuroimaging, Reproducibility
• Scientific‌​‌ Description:
  As deep learning​​ faces a reproducibility crisis​​​‌ and studies on deep‌ learning applied to neuroimaging‌​‌ are contaminated by methodological​​ flaws, there is an​​​‌ urgent need to provide‌ a safe environment for‌​‌ deep learning users to​​ help them avoid common​​​‌ pitfalls that will bias‌ and discredit their results.‌​‌ Several tools have been​​ proposed to help deep​​​‌ learning users design their‌ framework for neuroimaging data‌​‌ sets. ClinicaDL has been​​ developed to bring answers​​​‌ to three common issues‌ encountered by deep learning‌​‌ users who are not​​ always familiar with neuroimaging​​​‌ data: (1) the format‌ and preprocessing of neuroimaging‌​‌ data sets, (2) the​​ contamination of the evaluation​​​‌ procedure by data leakage‌ and (3) a lack‌​‌ of reproducibility. The combination​​ of ClinicaDL and its​​​‌ companion project Clinica allows‌ performing an end-to-end neuroimaging‌​‌ analysis, from the download​​ of raw data sets​​​‌ to the interpretation of‌ trained networks, including neuroimaging‌​‌ preprocessing, quality check, label​​ definition, architecture search, and​​​‌ network training and evaluation.‌
• Functional Description:
  ClinicaDL is‌​‌ a Python open-source software​​ for neuroimaging data processing​​​‌ with deep learning. This‌ software includes many functionalities,‌​‌ such as neuroimaging preprocessing,​​ synthetic dataset generation, label​​​‌ definition, data split with‌ similar demographics, architecture search,‌​‌ network training, performance evaluation​​ and trained network interpretation.​​​‌ The three main objectives‌ of ClinicaDL are to‌​‌ (1) help manipulate neuroimaging​​​‌ data sets, (2) prevent​ data leakage from biasing​‌ results and (3) reproduce​​ deep learning experiments.
• URL:​​​‌
  http://­clinicadl.­aramislab.­fr/
• Publications:
  hal-03351976,​ hal-02562504, hal-04279014,​‌ hal-04419141
• Contact:
  Ninon Burgos​​
• Participants:
  Ninon Burgos, Olivier​​​‌ Colliot, Thibault De Varax,​ Maelys Solal, Hugues Roy,​‌ Camille Brianceau, Mauricio Diaz,​​ Ravi Hassanaly, Alexandre Routier,​​​‌ Elina Thibeau-Sutre
• Partners:
  Institut​ du Cerveau et de​‌ la Moelle épinière (ICM),​​ CNRS, INSERM, Sorbonne Université​​​‌

7.1.3 leaspy

• Name:
  Learning​ spatiotemporal patterns in python​‌
• Keywords:
  Clinical analysis, Medical​​ applications, Personalized medicine
• Functional​​​‌ Description:
  Leaspy, standing for​ LEArning Spatiotemporal Patterns in​‌ Python, has been developed​​ to analyze longitudinal (or​​​‌ sequential) data that correspond​ to the measurements of​‌ a long-term progression. Said​​ differently, each sequence of​​​‌ repeated observations derives from​ a portion of the​‌ global process, with a​​ certain variability between sequence.​​​‌
• Release Contributions:

  1.The new​ structure is a result​‌ of a global refactoring:​​ o This breaking change​​​‌ is the result of​ a major refactoring of​‌ how model parameters and​​ variables are handled. o​​​‌ The new architecture is​ more modular and is​‌ designed to simplify the​​ definition and extension of​​​‌ models compared to v1.​ o The codebase has​‌ been structured to closely​​ mirror the mathematical formulation​​​‌ of the models. o​ At its core, distinct​‌ classes are implemented, following​​ a well-defined hierarchy of​​​‌ dependencies and inheritance, for​ the model and the​‌ algorithm class, using a​​ Directed Acyclic Graph (DAG)​​​‌ and a 'family' concept.​ o This modular and​‌ transparent architecture ensures clarity,​​ extensibility, and consistency, while​​​‌ its straightforward structure greatly​ simplifies the development and​‌ integration of new model​​ or algorithm variants. o​​​‌ A State builds on​ this structure: it is​‌ a DAG with an​​ additional mapping between node​​​‌ names and their current​ values, effectively holding both​‌ the model's blueprint and​​ the values currently loaded.​​​‌

  2.New models are added:​ o joint: for the​‌ joint modelisation of patient​​ outcomes and events o​​​‌ mixture: for the clustering​ of patients with similar​‌ spatiotemporal profiles

  3.Documentation is​​ updated and completed. The​​​‌ documentation page that comes​ with this release provides​‌ a detailed description of​​ the mathematical intuition behind​​​‌ leaspy, the models and​ the algorithms implemented and​‌ the interpretation of the​​ results. It also comes​​​‌ with an example gallery​ with complete notebooks fitting​‌ different models with synthetic​​ data.

• Publications:
  hal-01540367,​​​‌ hal-01964821, hal-03877293,​ hal-04319442, hal-04023781,​‌ hal-03831598, tel-04770912,​​ hal-04095450, hal-04216957,​​​‌ hal-04848392
• Contact:
  Sophie Tezenas​ Du Montcel
• Participants:
  Lea​‌ Aguilhon, Gabrielle Casimiro, Raphaël​​ Couronné, Némo Fournier, Nicolas​​​‌ Gensollen, Sofia Kaisaridi, Igor​ Koval, Etienne Maheux, Sebastian​‌ Mendez Pineda, Juliette Ortholand,​​ Pierre-Emmanuel Poulet, Maylis Tran,​​​‌ Caglayan Tuna, Arnaud Valladier​
• Partners:
  Université Paris Cité,​‌ INSERM

7.1.4 sndsTools

• Name:​​
  Extraction of healthcare utilisation​​​‌ data from the SNDS​ with R
• Keywords:
  Electronic​‌ Medical Records, Real world​​ data
• Functional Description:
  sndsTools​​​‌ is a community-driven R​ package designed to streamline​‌ the extraction, cleaning, and​​ harmonisation of healthcare utilisation​​​‌ data from the French​ National Health Data System​‌ (SNDS), which is a​​ crucial resource for large-scale​​ real-world evidence research. Initiated​​​‌ after multidisciplinary discussions at‌ the Emois congress in‌​‌ March 2024, and coordinated​​ by T. Nedelec (ARAMIS)​​​‌ alongside collaborators from the‌ French Health Authority (HAS)‌​‌ and AP-HM, sndsTools addresses​​ core challenges in SNDS​​​‌ data management. The package‌ greatly simplifies recurring steps‌​‌ in SNDS-based studies, making​​ routine data extraction and​​​‌ reproducibility accessible for a‌ growing community of users‌​‌ in epidemiology and medical​​ data science. With a​​​‌ modular, evolving design and‌ long-term support ambitions, sndsTools‌​‌ is released under the​​ EUPL licence to foster​​​‌ open, collaborative development and‌ ensure compatibility with the‌​‌ wider public health research​​ ecosystem.
• URL:
  https://­sndstoolers.­github.­io/­sndsTools/
• Contact:​​​‌
  Thomas Nedelec
• Participants:
  Thomas‌ Nedelec, Antoine Belloir, Marc‌​‌ Dibling, Matthieu Doutreligne, Leo​​ Guillon, Thomas Soeiro

7.1.5​​​‌ PhagoStat

• Name:
  Efficient quantification‌ of cell phagocytosis in‌​‌ neurodegenerative disease studies
• Keywords:​​
  Explainable Artificial Intelligence, Deep​​​‌ learning, Scalability, Live-cell microscopy,‌ Microscopy
• Scientific Description:
  This‌​‌ pipeline is an integrated,​​ end-to-end solution for data-sequence​​​‌ handling, video-based analysis, noise‌ management, quantitative analysis, and‌​‌ statistical reporting. To the​​ best of our knowledge,​​​‌ unique in its scope‌ and functionality, we have‌​‌ made this innovative tool​​ publicly available on GitHub.​​​‌ As an added feature,‌ especially beneficial for less‌​‌ technical users, we included​​ a pre-coded, user-friendly UX​​​‌ and a framework for‌ an HPC environment, it‌​‌ operates with a single​​ command line. User-friendly UX​​​‌ and HPC support components‌ are optional for highly‌​‌ technical users. Given that​​ the source code for​​​‌ all modules, along with‌ the UX and HPC‌​‌ framework, is publicly accessible,​​ it allows for usage,​​​‌ modification, and potential enhancements‌ by the community. Applying‌​‌ the PhagoStat pipeline to​​ microglial cells has yielded​​​‌ statistically significant findings. Our‌ discovery that Frontotemporal Dementia‌​‌ (FTD) mutant cells exhibit​​ increased size and activity​​​‌ compared to wild-type cells‌ is a novel insight,‌​‌ contributing significantly to our​​ understanding of neurodegenerative diseases​​​‌ and potentially catalyzing further‌ research in this domain.‌​‌
• Functional Description:
  The PhagoStat​​ pipeline is able to​​​‌ process large data-sets and‌ includes a data quality‌​‌ verification module to counteract​​ potential perturbations such as​​​‌ microscope movements and frame‌ blurring. We also propose‌​‌ an explainable cell segmentation​​ module to improve the​​​‌ interpretability of deep learning‌ methods compared to black-box‌​‌ algorithms. This includes two​​ interpretable deep learning capabilities:​​​‌ visual explanation and model‌ simplification. We demonstrate that‌​‌ interpretability in deep learning​​ is not the opposite​​​‌ of high performance, by‌ additionally providing essential deep‌​‌ learning algorithm optimization insights​​ and solutions. Besides, incorporating​​​‌ interpretable modules results in‌ an efficient architecture design‌​‌ and optimized execution time.​​ We apply this pipeline​​​‌ to quantify and analyze‌ microglial cell phagocytosis in‌​‌ frontotemporal dementia (FTD) and​​ obtain statistically reliable results​​​‌ showing that FTD mutant‌ cells are larger and‌​‌ more aggressive than control​​ cells. The method has​​​‌ been tested and validated‌ on several public benchmarks‌​‌ by generating state-of-the art​​ performances. To stimulate translational​​​‌ approaches and future studies,‌ we release an open-source‌​‌ end-to-end pipeline and a​​ unique microglial cells phagocytosis​​​‌ dataset for immune system‌ characterization in neurodegenerative diseases‌​‌ research. This pipeline and​​​‌ the associated dataset will​ consistently crystallize future advances​‌ in this field, promoting​​ the development of efficient​​​‌ and effective interpretable algorithms​ dedicated to the critical​‌ domain of neurodegenerative diseases'​​ characterization.
• Release Contributions:
  Initial​​​‌ version 1.0
• URL:
  https://­github.­com/­ounissimehdi/­PhagoStat​
• Publication:
  hal-04067345v3
• Contact:
  Daniel​‌ Racoceanu
• Participants:
  Daniel Racoceanu,​​ Medhi Ounissi

ADNP-15

• Contributors:
  Chenxi​ Zhao, Jianqiang Li, Qing​‌ Zhao, Jing Bai, Susana​​ Boluda, Benoit Delatour, Lev​​​‌ Stimmer, Daniel Racoceanu ,​ Gabriel Jimenez, Guanghui Fu​‌
• Description:
  The ADNP-15 dataset​​ 88 represents a unique​​​‌ open-source resource for the​ quantitative analysis of Alzheimer's​‌ disease histopathology. Unlike existing​​ collections, it provides expertly​​​‌ annotated whole-slide images of​ neuritic plaques, lesions that​‌ combine amyloid deposits with​​ surrounding tau-positive dystrophic neurites,​​​‌ captured across realistic staining​ and acquisition conditions. Its​‌ dual focus on lesion​​ segmentation and stain variability​​​‌ makes it the first​ dataset specifically designed to​‌ benchmark both deep-learning architectures​​ and stain normalisation methods​​​‌ for AD pathology. By​ releasing all images, annotations,​‌ and code openly, ADNP-15​​ establishes a reproducible foundation​​​‌ for developing and comparing​ algorithms, filling a critical​‌ gap in large-scale, high-quality​​ data for computational neuropathology.​​​‌
• Dataset DOI:
  doi:10.5281/zenodo.15009434
• Publication:​
  Chenxi Zhao, Jianqiang Li,​‌ Qing Zhao, Jing Bai,​​ Susana Boluda, et al..​​​‌ ADNP-15: An Open-Source Histopathological​ Dataset for Neuritic Plaque​‌ Segmentation in Human Brain​​ Whole Slide Images with​​​‌ Frequency Domain Image Enhancement​ for Stain Normalization. Innovation​‌ and Research in BioMedical​​ engineering, 2025, 46 (6),​​​‌ pp.100913. ⟨10.1016/j.irbm.2025.100913⟩.​ ⟨hal-05286275⟩

10.1.1 Inria associate​​​‌ team not involved in​ an IIL or an​‌ international program

10.2.1 Horizon Europe​‌

10.2.2 Other European programs/initiatives​​

10.3.1 IHU

10.3.2 3IA Institutes &‌ IA-Clusters

10.3.3 ANR

10.3.4 PEPR

10.3.5 RHU

10.3.6 Other national programs‌

11.1.1 Scientific events: selection​​​‌

11.1.2 Journal​‌

11.1.3 Invited talks

• Olivier​​​‌ Colliot was invited to‌ give a talk at‌​‌ the Annual French Conference​​ on Artificial Intelligence for​​​‌ Biomedical Imaging (IABM), Nice,‌ France, 2025.
• Olivier Colliot‌​‌ was invited to give​​ a talk at the​​​‌ Indo-French Dialogue on AI‌ in Healthcare, Paris, France,‌​‌ 2025.
• Olivier Colliot was​​ invited to give a​​​‌ talk at MICCAI webinar‌ series, online, 2025.
• Ninon‌​‌ Burgos was invited to​​ give a talk at​​​‌ the AI4Health Summer School‌ (Paris, France).
• Ninon Burgos‌​‌ was invited to give​​ a talk at the​​​‌ Deep Learning for Medical‌ Imaging School (Lyon, France).‌​‌
• Ninon Burgos was invited​​ to give a talk​​​‌ at the Journée des‌ ingénieurs du Programme National‌​‌ de Recherche en Intelligence​​ Artificielle (Paris, France).
• Ninon​​​‌ Burgos was invited to‌ give a talk at‌​‌ the Journées scientifiques Inria​​ handicap et numérique (Paris,​​​‌ France).
• Ninon Burgos was‌ invited to give a‌​‌ talk at Radiologie Aujourd'hui​​ et Demain (Angers, France).​​​‌
• Ninon Burgos was invited‌ to give a talk‌​‌ at the Symposium Citadel​​ Franco-Québecois (Montreal, Canada).
• Ninon​​​‌ Burgos was invited to‌ give seminars in the‌​‌ context of the CONNExIN​​ (COmprehensive Neuroimaging aNalysis Experience​​​‌ In resource constraiNed settings)‌ programme (online), COMPASS (COnnecting‌​‌ Minds to Progress AI​​ in Medicine Seminar Series)​​​‌ (online) and the McConnell‌ Brain Imaging Centre seminar‌​‌ (Montreal, Canada).
• Sophie Tezenas​​ Du Montcel was invited​​​‌ to give a talk‌ at the meeting Human‌​‌ trajectories: models and applications​​ (Paris, France).
• Daniel Racoceanu​​​‌ was invited speaker at‌ the Journées Ouvertes en‌​‌ Biologie, Informatique et Mathématiques​​ (JOBIM), Bordeaux, France.
• Daniel​​​‌ Racoceanu was invited spaker‌ at the Computer Vision‌​‌ for Drug Discovery (CVDD)​​ workshop, IEEE/CVF Conference on​​​‌ Computer Vision and Pattern‌ Recognition (CVPR), Nashville, TN,‌​‌ USA.

11.1.4 Leadership within​​ the scientific community

• Olivier​​​‌ Colliot is founding member‌ of the board of‌​‌ IABM (The French Society​​ for Artificial Intelligence in​​​‌ Biomedical Imaging) (since 2025).‌
• Olivier Colliot is a‌​‌ member of the board​​ and the treasurer of​​​‌ the MICCAI Special Interest‌ Group on Challenges in‌​‌ AI (since 2024).
• Daniel​​ Racoceanu is a member​​​‌ of the Advisory Board‌ of the European Society‌​‌ of Integrative Digital Pathology​​ (ESDIP).
• Sophie Tezenas Du​​​‌ Montcel is a member‌ of the Critical Path‌​‌ to Therapeutics for the​​ Ataxias.

11.1.5 Scientific expertise​​​‌

• Ninon Burgos reviewed grant‌ applications for the Fonds‌​‌ de recherche du Québec,​​​‌ MIAI Cluster Chair, KU​ Leuven Industrial Research Fund​‌ Council and ANR PRCI.​​
• Ninon Burgos was a​​​‌ member of the committee​ for the Inserm/IReSP MESSIDORE​‌ Programme (2025).
• Ninon Burgos​​ was a member of​​​‌ the jury for the​ Health Data Hub Data​‌ Challenges en Santé.
• Ninon​​ Burgos is a member​​​‌ of the Scientific and​ Ethical Committee of the​‌ Paris university hospital trust's​​ clinical data warehouse (EDS​​​‌ AP-HP) (2024–).
• Ninon Burgos​ was a member of​‌ the jury for the​​ recruitment of PR[AI]RIE-PSAI Fellows​​​‌ in Artificial Intelligence.
• Daniel​ Racoceanu is elected member​‌ of the Inria Evaluation​​ Committee (Inria CE) for​​​‌ the quadriennal mandate 2023–2027.​
• Daniel Racoceanu was a​‌ member of the Scientific​​ Evaluation Committee “Interfaces: mathematics,​​​‌ digital sciences - biology,​ health” (CE45) of the​‌ French National Research Agency​​ ANR (2025).
• Daniel Racoceanu​​​‌ participated, as reviewer to​ the recruitment jury for​‌ the Junior Professorship (CPJ)​​ position at MINES Paris​​​‌ - PSL.
• Daniel Racoceanu​ participated, as reviewer to​‌ the recruitment jury for​​ the Junior Professorship (CPJ)​​​‌ position at the University​ of Montpellier.
• Daniel Racoceanu​‌ participated, as reviewer to​​ the recruitment jury for​​​‌ the Full Professor position​ at the Sorbonne University.​‌
• Daniel Racoceanu participated, as​​ reviewer to the recruitment​​​‌ jury for the Full​ Professor position at Polytech​‌ Sorbonne engineering school.
• Sophie​​ Tezenas Du Montcel is​​​‌ member of the Ataxia​ Advisory Committee for Therapeutics​‌ (ACT of Ataxia Global​​ Initiative).
• Sophie Tezenas Du​​​‌ Montcel is a member​ of the scientific board​‌ of the National Bank​​ for Rare Diseases (Banque​​​‌ Nationale de Données Maladies​ Rares, BNDMR).
• Sophie Tezenas​‌ Du Montcel is a​​ member of scientific board​​​‌ of the CERMOI study.​

11.1.6 Research administration

• Olivier​‌ Colliot is Deputy Scientific​​ Director of the ICM​​​‌ (since 2025).
• Olivier Colliot​ is Inaugural Director of​‌ the Centre for Artificial​​ Intelligence and Data Science​​​‌ of the ICM (since​ 2025).
• Olivier Colliot is​‌ a member of the​​ Executive Committee (CODIR) of​​​‌ the ICM (since 2025).​
• Olivier Colliot is a​‌ member of the “Bureau​​ du Comité des Projets”​​​‌ of the Inria Paris​ Centre.

11.1.7 Research committees​‌

• Sophie Tezenas Du Montcel​​ is a member of​​​‌ the bureau of the​ Conseil national des universités​‌ 4604.
• Ninon Burgos is​​ the scientific secretary of​​​‌ the Institute Scientific Board​ of CNRS Informatics.

11.2.1 Teaching​

11.2.2 Supervision​‌

• PhD in progress: Charles​​ Heitz , “Deep learning​​​‌ for assisting clinical decisions​ in brain imaging: trustworthy​‌ validation and benchmarking”, started​​ in 2025, supervisors: Olivier​​​‌ Colliot
• PhD in progress:​ Pascaline Andre , “Statistical​‌ evaluation of models and​​ machine learning procedures in​​​‌ medical imaging”, started in​ 2024, supervisors: Olivier Colliot​‌ and Sophie Tezenas du​​ Montcel
• PhD completed in​​​‌ 2025: Guanghui Fu ,​ “Segmentation, classification and generative​‌ models for computer-aided diagnosis​​ of neurological diseases from​​​‌ neuroimaging data”, started in​ 2021, supervisors: Olivier Colliot​‌ and Didier Dormont
• PhD​​ completed in 2025: Arya​​​‌ Yazdan-Panah , “Deep learning​ for multimodal image analysis​‌ in multiple sclerosis”, started​​ in 2021, supervisors: Olivier​​​‌ Colliot and Bruno Stankoff​
• PhD in progress: Manon​‌ Heffernan , “Artificial intelligence​​ tools for clinical data​​​‌ warehouses in neuroimaging”, started​ in 2024, supervisors: Olivier​‌ Colliot and Ninon Burgos​​
• PhD in progress: Matthieu​​​‌ Joulot , “Longitudinal processing​ of multimodal brain imaging​‌ for the study of​​ neurodegenerative diseases”, started in​​​‌ 2024, supervisors: Ninon Burgos​ and Olivier Colliot
• PhD​‌ in progress: Maëlys Solal​​ , “Robust anomaly detection​​​‌ in multimodal neuroimaging”, started​ in 2023, supervisor: Ninon​‌ Burgos
• PhD in progress:​​ Hugues Roy , “Pseudo-healthy​​​‌ image synthesis for the​ detection of anomalies in​‌ the brain, a multi-modal​​ approach”, started in 2024,​​​‌ supervisor: Ninon Burgos
• PhD​ completed in 2025: Élise​‌ Delzant, “Methods for big-data​​ neuroimaging analyses”, started in​​​‌ 2022, supervisors: Baptiste Couvy-Duchesne​ and Olivier Colliot
• PhD​‌ in progress: Maylis Tran​​ , “Optimisation du design​​​‌ d'essai clinique à l'aide​ de données d'histoire naturelle”,​‌ started in 2024, supervisor:​​ Sophie Tezenas Du Montcel​​​‌
• PhD in progress: Marc​ Dibling , “Parcours de​‌ soin des patients atteints​​ de maladies neurodégénératives rares”,​​​‌ started in 2023, supervisor:​ Sophie Tezenas Du Montcel​‌
• PhD in progress: Sofia​​ Kaisaridi , “Modélisation multimarqueurs​​​‌ de l'évolution clinique et​ en imagerie cérébrale de​‌ patients CADASIL et de​​ son influence sur un​​​‌ évènement censure”, started in​ 2022, supervisor: Sophie Tezenas​‌ Du Montcel
• PhD in​​ progress: Ayse Gungor ,​​ “Correlation between eye and​​​‌ brain pathologies”, started in‌ 2023, supervisors: Dan Milea‌​‌ and Daniel Racoceanu
• PhD​​ in progress: Ilias Sarbout​​​‌ , “Artificial Vision by‌ fMRI analysis and XAI‌​‌ approaches", started in 2023,​​ supervisors: Dan Milea and​​​‌ Daniel Racoceanu
• PhD in‌ progress: Esther Kozlowski ,‌​‌ “A responsible artificial intelligence​​ framework for modeling the​​​‌ progression of Parkinson's disease”,‌ started in 2023, supervisors:‌​‌ Marie Vidailhet and Daniel​​ Racoceanu
• PhD in progress:​​​‌ Swann Ruyter , “ComPath:‌ Next Generation Computational Pathomics‌​‌ for Personalized Medicine. Explainable​​ Deep Learning Integration of​​​‌ Computational Pathology and Spatial‌ Transcriptomics”, started in 2024,‌​‌ supervisor: Daniel Racoceanu
• PhD​​ in progress: Mehdi Hamadache​​​‌ , “Physics Informed AI‌ meets Mechanistic Modeling: Predicting‌​‌ Cancer Immunotherapy Outcomes using​​ Multimodal Computational Pathology”', started​​​‌ in 2025, supervisor: Daniel‌ Racoceanu
• PhD completed in‌​‌ 2025: Octave Guinebretiere ,​​ “Early prediction of neurodegenerative​​​‌ diseases using large transnational‌ electronic health records databases‌​‌ for better prevention”, started​​ in 2022, supervisors: Stanley​​​‌ Durrleman and Thomas Nedelec‌

11.2.3 Juries

• Ninon Burgos‌​‌ participated, as reviewer, to​​ the PhD committee of​​​‌ Juliette Moreau, Université Claude‌ Bernard Lyon 1.
• Ninon‌​‌ Burgos participated, as reviewer,​​ to the PhD committee​​​‌ of Ashay Patel, King's‌ College London.
• Ninon Burgos‌​‌ participated, as reviewer, to​​ the PhD committee of​​​‌ Aghiles Kebaili, Université de‌ Rouen.
• Ninon Burgos participated,‌​‌ as reviewer, to the​​ PhD committee of Élodie​​​‌ Piot, Université Grenoble Alpes.‌
• Ninon Burgos participated to‌​‌ the individual PhD student​​ monitoring committee of Florencia​​​‌ Boccarato, Université Côte d'Azur.‌
• Ninon Burgos participated to‌​‌ the individual PhD student​​ monitoring committee of Baptiste​​​‌ Pierrard, Université de Lyon.‌
• Ninon Burgos participated to‌​‌ the individual PhD student​​ monitoring committee of Daniele​​​‌ Falcetta, EURECOM .
• Ninon‌ Burgos participated to the‌​‌ individual PhD student monitoring​​ committee of Trang Nguyen,​​​‌ Université Côte d'Azur .‌
• Ninon Burgos participated to‌​‌ the individual PhD student​​ monitoring committee of Franklin​​​‌ Sierra, Institut Polytechnique de‌ Paris, France & Universidad‌​‌ Industrial de Santander, Colombia.​​
• Daniel Racoceanu participated, as​​​‌ reviewer, to the PhD‌ committee of Alexandre Martin,‌​‌ Université Côte d'Azur.
• Daniel​​ Racoceanu participated, as president,​​​‌ to the PhD committee‌ of Victorien Quevit, University‌​‌ of Rennes.
• Daniel Racoceanu​​ participated to the individual​​​‌ PhD student monitoring committee‌ of Marie Arrivat, Institut‌​‌ Polytechnique de Paris.
• Daniel​​ Racoceanu participated to the​​​‌ individual PhD student monitoring‌ committee of Tiziana Tocci,‌​‌ Institut Curie, Paris.
• Daniel​​ Racoceanu participated to the​​​‌ individual PhD student monitoring‌ committee of Paul Barthe,‌​‌ University of Caen.
• Olivier​​ Colliot participated, as president,​​​‌ to the PhD committee‌ of Marianne Golse, Sorbonne‌​‌ University.
• Sophie Tezenas Du​​ Montcel participated, as reviewer,​​​‌ to the PhD committee‌ of Niels Hendrickx, Université‌​‌ Paris Cité.
• Sophie Tezenas​​ Du Montcel participated, as​​​‌ reviewer, to the PhD‌ committee of Théo Silvestre,‌​‌ Université Paris Saclay.

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

• Olivier Colliot contributed‌​‌ to the popular science​​ book “Tout comprendre (ou​​​‌ presque) sur l'intelligence artificielle”‌ (CNRS éditions).

International journals‌

• 63 articleR.Remy‌​‌ Ben Messaoud, V.​​ L.Vincent Le Du​​​‌, C.Camile Bousfiha‌, M.-C.Marie-Constance Corsi‌​‌, J.Juliana Gonzalez-Astudillo​​, B. C.Brigitte​​​‌ Charlotte Kaufmann, T.‌Tristan Venot, B.‌​‌Baptiste Couvy-Duchesne, L.​​Lara Migliaccio, C.​​​‌Charlotte Rosso, P.‌Paolo Bartolomeo, M.‌​‌Mario Chavez and F.​​Fabrizio de Vico Fallani​​​‌. Low-dimensional controllability of‌ brain networks.PLoS‌​‌ Computational Biology2025HAL​​
• 64 articleQ.Quentin​​​‌ Calonge, O.Octave‌ Guinebretiere, T.Thomas‌​‌ Nedelec, A.Aurélie​​ Hanin, F.Francois​​​‌ Le Gac, M.‌Mario Chavez, F.‌​‌Florence Tubach, S.​​Sophie Tezenas Du Montcel​​​‌ and V.Vincent Navarro‌. Recurrence and Mortality‌​‌ After a First Status​​ Epilepticus.Neurology104​​​‌12June 2025HAL‌DOI
• 65 articleQ.‌​‌Quentin Calonge, A.​​Aurélie Hanin, E.​​​‌Edouard Januel, O.‌Octave Guinebretiere, T.‌​‌Thomas Nedelec, F.​​Francois Le Gac,​​​‌ M.Mario Chavez,‌ S.Sophie Tezenas Du‌​‌ Montcel and V.Vincent​​ Navarro. Incidence, mortality,​​​‌ and management of status‌ epilepticus from 2012 to‌​‌ 2022: An 11‐year nationwide​​ study.EpilepsiaSeptember​​​‌ 2025HALDOI
• 66‌ articleQ.Quentin Calonge‌​‌, V.Vincent Navarro​​ and S.Sophie Tezenas​​​‌ Du Montcel. How‌ Reliable Is the G41‌​‌ Discharge Code for Status​​ Epilepticus?Brain and Behavior​​​‌1532025,‌ e70443HALDOIback‌​‌ to text
• 67 article​​C.-H.Chih-Hao Chen,​​​‌ Y.-W.Yu-Wen Cheng,‌ R.Ruiting Zhang,‌​‌ S.Sophie Tezenas Du​​ Montcel, S.Stéphanie​​​‌ Guey, D.Dominique‌ Hervé, S.-C.Sung-Chun‌​‌ Tang and H.Hugues​​ Chabriat. Intracerebral Hemorrhage​​​‌ in Patients With CADASIL:‌ Additive Impact of the‌​‌ NOTCH3 R544C Variant and​​​‌ Hypertension?Stroke568​August 2025, 2159-2166​‌HALDOI
• 68 article​​E.Elise Delzant,​​​‌ O.Olivier Colliot and​ B.Baptiste Couvy-Duchesne.​‌ Choice of processing pipelines​​ for T1-weighted brain MRI​​​‌ impacts association and prediction​ analyses.Human Brain​‌ Mapping4616October​​ 2025HALDOIback​​​‌ to textback to​ text
• 69 articleC.​‌Cécile Di Folco,​​ C.Charlotte Dubec‐fleury,​​​‌ A.Andreas Traschütz,​ C.Christoph Kessler,​‌ S.Selina Reich,​​ C.Cynthia Gagnon,​​​‌ I.Isabelle Lessard,​ X.Xavier Rodrigue,​‌ S.Sirio Cocozza,​​ S.Sara Satolli,​​​‌ F. M.Filippo M​ Santorelli, A.Alexandra​‌ Durr, A.Anna​​ Heinzmann, B. P.​​​‌Bart P van de​ Warrenburg, I. H.​‌Ilse H.J. Willemse,​​ A. N.A. Nazli​​​‌ Başak, A.Atay​ Vural, B.Bernard​‌ Brais, S.Stephan​​ Klebe, R.Rita​​​‌ Horvath, R.Rebecca​ Schüle, M.Matthis​‌ Synofzik and S.Sophie​​ Tezenas Du Montcel.​​​‌ Spastic Ataxia Composite (​ SPAXCOM ): A Scale​‌ to Evaluate the Progression​​ of Subjects with Spasticity​​​‌ and Ataxia.Movement​ DisordersAugust 2025HAL​‌DOIback to text​​
• 70 articleR.Rosana​​​‌ El Jurdi, G.​Gaël Varoquaux and O.​‌Olivier Colliot. Confidence​​ intervals for performance estimates​​​‌ in brain MRI segmentation​.Medical Image Analysis​‌103July 2025,​​ 103565HALDOIback​​​‌ to textback to​ text
• 71 articleG.​‌Guanghui Fu, L.​​Lucia Nichelli, D.​​​‌Darío Herrán de la​ Gala, S.Sophie​‌ Loizillon, C.Camile​​ Bousfiha, R.Romain​​​‌ Valabregue, A.Agusti​ Alentorn, K.Khê​‌ Hoang-Xuan, B.Bertrand​​ Mathon, C.Carole​​​‌ Soussain, J. P.​Jean Pierre Marolleau,​‌ J.Jérôme Paillassa,​​ L.Luc Taillandier,​​​‌ P.Philippe Agapé,​ A.Anna Schmitt,​‌ O.Olivier Chinot,​​ G.Guido Ahle,​​​‌ D.Didier Dormont,​ C.Caroline Houillier,​‌ S.Stéphane Lehéricy and​​ O.Olivier Colliot.​​​‌ Automatic Segmentation of Primary​ Central Nervous System Lymphoma​‌ at Clinical Routine Postcontrast​​ T1-weighted MRI.Radiology:​​​‌ Imaging Cancer75​September 2025HALDOI​‌back to textback​​ to textback to​​​‌ text
• 72 articleA.​Ayse Gungor, I.​‌Ilias Sarbout, A.​​Aubrey Gilbert, S.​​​‌Steffen Hamann, P.​Pierre Lebranchu, C.​‌Cristina Hobeanu, P.​​Philippe Gohier, C.​​​‌Catherine Vignal-Clermont, O.​Oana Dumitrascu, S.​‌Salomon‐yves Cohen, W.​​Wolf Lagrèze, N.​​​‌Nicolas Feltgen, F.​Frank van der Heide​‌, C.Cédric Lamirel​​, J.Jost Jonas​​​‌, M.Michael Obadia​, D.Daniel Racoceanu​‌ and D.Dan Milea​​. Artificial Intelligence‐Based Detection​​​‌ of Central Retinal Artery​ Occlusion Within 4.5 Hours​‌ on Standard Fundus Photographs​​.Journal of the​​​‌ American Heart Association14​13June 2025HAL​‌DOIback to text​​
• 73 articleR.Ravi​​​‌ Hassanaly, M.Maëlys​ Solal, O.Olivier​‌ Colliot and N.Ninon​​ Burgos. Benchmarking 3D​​ generative autoencoders for pseudo-healthy​​​‌ reconstruction of brain 18F-fluorodeoxyglucose‌ positron emission tomography.‌​‌Journal of Medical Imaging​​125October 2025​​​‌, 054005HALDOI‌back to text
• 74‌​‌ articleS.Sofia Kaisaridi​​, D.Dominique Herve​​​‌, A.Aude Jabouley‌, S.Sonia Reyes‌​‌, C.Carla Machado​​, S.Stéphanie Guey​​​‌, A.Abbas Taleb‌, F.Fanny Fernandes‌​‌, H.Hugues Chabriat​​ and S.Sophie Tezenas​​​‌ Du Montcel. Determining‌ Clinical Disease Progression in‌​‌ Symptomatic Patients With CADASIL​​.Neurology1041​​​‌January 2025HALDOI‌back to textback‌​‌ to textback to​​ text
• 75 articleS.​​​‌Sophie Loizillon, S.‌Simona Bottani, A.‌​‌Aurélien Maire, S.​​Sebastian Ströer, L.​​​‌Lydia Chougar, D.‌Didier Dormont, O.‌​‌Olivier Colliot and N.​​Ninon Burgos. Automatic​​​‌ quality control of brain‌ 3D FLAIR MRIs for‌​‌ a clinical data warehouse​​.Medical Image Analysis​​​‌103July 2025,‌ 103617HALDOIback‌​‌ to textback to​​ text
• 76 articleL.​​​‌ E.Laura E Marin‌, D. I.Daniel‌​‌ I Zavaleta-Guzman, J.​​ I.Jessyca I Gutierrez-Garcia​​​‌, D.Daniel Racoceanu‌ and F. L.Fanny‌​‌ L Casado. Prediction​​ of biochemical prostate cancer​​​‌ recurrence from any Gleason‌ score using robust tissue‌​‌ structure and clinically available​​ information.Discover Oncology​​​‌161February 2025‌, 128HALDOI‌​‌back to text
• 77​​ articleJ.Juliette Ortholand​​​‌, N.Nicolas Gensollen‌, S.Stanley Durrleman‌​‌ and S.Sophie Tezenas​​ Du Montcel. Joint​​​‌ model with latent disease‌ age: overcoming the need‌​‌ for reference time.​​Statistical Methods in Medical​​​‌ Research2025. In‌ press. HALDOI
• 78‌​‌ articleM.Mehdi Ounissi​​, I.Ilias Sarbout​​​‌, J.-P.Jean-Pierre Hugot‌, C.Christine Martinez-Vinson‌​‌, D.Dominique Berrebi​​ and D.Daniel Racoceanu​​​‌. Scalable, trustworthy generative‌ model for virtual multi-staining‌​‌ from H&E whole slide​​ images.PLoS Computational​​​‌ Biology2110October‌ 2025, e1013516HAL‌​‌DOIback to text​​
• 79 articleE.Emilien​​​‌ Petit, G.Giulia‌ Coarelli, D.David‌​‌ Morgan, P.Paulina​​ Cunha, H.Hortense​​​‌ Hurmic, J.Jennifer‌ Faber, M.Marcus‌​‌ Grobe-Einsler, T.Thiago​​ Rezende, S.-H.Sheng-Han​​​‌ Kuo, G.George‌ Wilmot, L.Liana‌​‌ Rosenthal, J.Jeremy​​ Schmahmann, T.Talene​​​‌ Yacoubian, S.Susan‌ Perlman, M.Michael‌​‌ Geschwind, C.Christopher​​ Gomez, T.Trevor​​​‌ Hawkins, S.S‌ Subramony, V. G.‌​‌Vikram G Shakkottai,​​ K.Khalaf Bushara,​​​‌ T.Theresa Zesiewicz,‌ S. M.Stefan M‌​‌ Pulst, Y.Young​​ Woo-Park, C.Christophe​​​‌ Lenglet, T.Thomas‌ Klockgether, H.Henry‌​‌ Paulson, A.Alexandra​​ Durr, G.Gülin​​​‌ Öz, T.Tetsuo‌ Ashizawa and S.Sophie‌​‌ Tezenas Du Montcel.​​ Predictive models for ataxia​​​‌ progression and conversion in‌ spinocerebellar ataxia type 1‌​‌ and 3.Brain​​ - A Journal of​​​‌ Neurology October 2025HAL‌DOIback to text‌​‌
• 80 articleE.Emilien​​​‌ Petit, D.Daniel​ López Domínguez, C.​‌Cecilia Marelli, S.​​Sabrina Sayah, S.​​​‌ M.Stefan M Pulst​, J.Jennifer Faber​‌, G.Gulin Oz​​, H. L.Henry​​​‌ L Paulson, T.​Tetsuo Ashizawa, S.​‌Sophie Tezenas Du Montcel​​, A.Alexandra Durr​​​‌, G.Giulia Coarelli​, C.Claire Ewenczyk​‌, A.Anna Heinzmann​​, P.Pauline Lallemant​​​‌, S.Solveig Heide​, P.Perrine Charles​‌, P.Paulina Cunha​​, H.Hortense Hurmic​​​‌, T.Thomas Klockgether​, M.Marcus Grobe-Einsler​‌, D.Demet Oender​​, O.Okka Kimmich​​​‌, N.Nina Roy​, I.Ilaria Giordano​‌, V.Veronika Purrer​​, C.Carolin Miklitz​​​‌, C.Cornelia Mccormick​, T.Titilayo Olubajo​‌, M.Matthew Burns​​, T.Trevor Hawkins​​​‌, K.Khalaf Bushara​, S.Susan Perlman​‌, S. H.S​​ H Subramony, D.​​​‌David Morgan, B.​Brianna Jackman, L.​‌Lauren Seeberger, D.​​ S.Drew Scott Kern​​​‌, G.George Wilmot​, L.Laura Scorr​‌, L.Liana Rosenthal​​, C.Chiadi Onyike​​​‌, M.Michael Geschwind​, A.Alexandra Nelson​‌, C.Cameron Dietiker​​, A.Armen Moughamian​​​‌, S.-H.Sheng-Han Kuo​, V.Vikram Shakkottai​‌, C.Christopher Gomez​​, M.Mahesh Padmanaban​​​‌, T.Talene Yacoubian​, M.Marissa Dean​‌, J.Jeremy Schmahmann​​, P. C.Peggy​​​‌ C Nopoulos, A.​Annie Killoran, P.​‌Puneet Opal, T.​​Theresa Zesiewicz, S.​​​‌Sharon Sha, V.​Veronica Santini, J.​‌Jacinda Sampson, P.​​Peter Morrison, E.​​​‌Erika Augustine and A.​Alex Paciorkowski. Prevalence,​‌ Severity, and Progression of​​ Cerebellar Cognitive-Affective Syndrome in​​​‌ Patients With Spinocerebellar Ataxias​.Neurology1055​‌2025HALDOI
• 81​​ articleP.-E.Pierre-Emmanuel Poulet​​​‌, M.Maylis Tran​, S.Sophie Tezenas​‌ Du Montcel, B.​​Bruno Dubois, S.​​​‌Stanley Durrleman and B.​Bruno Jedynak. Prediction-powered​‌ Inference for Clinical Trials:​​ application to linear covariate​​​‌ adjustment.BMC Medical​ Research MethodologyAugust 2025​‌HALback to text​​back to text
• 82​​​‌ articleD.Dario Saracino​, L.Lorenzo Cipriano​‌, M.Marion Houot​​, G.Giorgia Querin​​​‌, D.Daisy Rinaldi​, A.Armelle Rametti-Lacroux​‌, D.David Wallon​​, E.Emmanuel Gerardin​​​‌, P.Philippe Couratier​, M.Marie‐paule Boncoeur​‌, T.Thibaud Lebouvier​​, O.Olivier Colliot​​​‌, P.Pierre‐françois Pradat​, R.Raffaella Migliaccio​‌ and I.Isabelle Le​​ Ber. Quantifying multimodal​​​‌ longitudinal brain changes in​ presymptomatic C9orf72 disease.​‌Alzheimer's & Dementia :​​ the Journal of the​​​‌ Alzheimer's Association2112​December 2025HALDOI​‌
• 83 articleI.Ilias​​ Sarbout, A.Ayse​​​‌ Gungor, M.Mehdi​ Ounissi, S.Samy​‌ Zaher, M.Maurice​​ Ptito, R.Ron​​​‌ Kupers, D.Daniel​ Racoceanu and D.Dan​‌ Milea. Visual Prostheses​​ in the Era of​​​‌ Artificial Intelligence Technology.​Eye and BrainVolume​‌ 176August 2025​​, 95-113HALDOI​​back to text
• 84​​​‌ articleT.T. Soulier‌, N.Ninon Burgos‌​‌, R.Ravi Hassanaly​​, M.M. Pitombeira​​​‌, M.Maëlys Solal‌, H.Hugues Roy‌​‌, M.M. Hamzaoui​​, A.A. Yazdan-Panah​​​‌, D.D. de‌ Paula Faria, C.‌​‌C. Louapre, B.​​B. Bodini, M.​​​‌M. Bottlaender, N.‌Nicholas Ayache, O.‌​‌Olivier Colliot and B.​​B. Stankoff. Artificial​​​‌ intelligence in presymptomatic neurological‌ diseases: Bridging normal variation‌​‌ and prodromal signatures.​​Revue Neurologique1819​​​‌November 2025, 944-950‌HALDOI
• 85 article‌​‌L. S.Laura S​​ van Velzen, L.​​​‌Lejla Colic, Z.‌Zuriel Ceja, M.‌​‌ R.Maria R Dauvermann​​, L. M.Luca​​​‌ M Villa, H.‌ S.Hannah S Savage‌​‌, Y. J.Yara​​ J Toenders, N.​​​‌Niousha Dehestani, A.‌ H.Alyssa H Zhu‌​‌, A. I.Adrian​​ I Campos, L.​​​‌ E.Lauren E Salminen‌, M.Martin Alda‌​‌, I.Ingrid Agartz​​, N.Nina Alexander​​​‌, R.Rosa Ayesa-Arriola‌, E. D.Elizabeth‌​‌ D Ballard, N.​​Nerisa Banaj, C.​​​‌Carlotta Barkhau, Z.‌Zeynep Başgöze, J.‌​‌Jochen Bauer, F.​​Francesco Benedetti, K.​​​‌Klaus Berger, B.‌Bianca Besteher, K.‌​‌Katharina Brosch, M.​​Manuel Canal-Rivero, S.​​​‌Simon Cervenka, R.‌Romain Colle, C.‌​‌ G.Colm G Connolly​​, E.Emmanuelle Corruble​​​‌, P.Philippe Courtet‌, B.Baptiste Couvy-Duchesne‌​‌, B.Benedicto Crespo-Facorro​​, K. R.Kathryn​​​‌ R Cullen, U.‌Udo Dannlowski, J.‌​‌Jeremy Deverdun, A.​​ M.Ana M Diaz-Zuluaga​​​‌, L. M.Lorielle‌ M F Dietze,‌​‌ J. W.Jennifer W​​ Evans, N.Negar​​​‌ Fani, K.Kira‌ Flinkenflügel, N. P.‌​‌Naomi P Friedman,​​ I. H.Ian H​​​‌ Gotlib, N. A.‌Nynke A Groenewold,‌​‌ D.Dominik Grotegerd,​​ T.Tomas Hajek,​​​‌ A. S.Alexander S‌ Hatoum, M.Marco‌​‌ Hermesdorf, I. B.​​Ian B Hickie,​​​‌ Y.Yoshiyuki Hirano,‌ T. C.Tiffany C‌​‌ Ho, Y.Yuki​​ Ikemizu, F.Frank​​​‌ Iorfino, J. C.‌Jonathan C Ipser,‌​‌ Y.Yuko Isobe,​​ A. P.Andrea P​​​‌ Jackowski, F.Fabrice‌ Jollant, T.Tilo‌​‌ Kircher, M.Melissa​​ Klug, S.-M.Sheri-Michelle​​​‌ Koopowitz, A.Anna‌ Kraus, A.Axel‌​‌ Krug, E.Emmanuelle​​ Le Bars, E.​​​‌ J.Elisabeth J Leehr‌, M.Meng Li‌​‌, E. T.Elizabeth​​ T C Lippard,​​​‌ C.Carlos Lopez-Jaramillo,‌ I. I.Ivan I‌​‌ Maximov, A. M.​​Andrew M Mcintosh,​​​‌ K. A.Katie A‌ Mclaughlin, S. R.‌​‌Sean R Mcwhinney,​​ S.Susanne Meinert,​​​‌ E.Elisa Melloni,‌ P. B.Philip B‌​‌ Mitchell, B.Benson​​ Mwangi, I.Igor​​​‌ Nenadić, S.Stener‌ Nerland, E.Emilie‌​‌ Olie, V.Victor​​ Ortiz-García de la Foz​​​‌, P. M.Pedro‌ M Pan, F.‌​‌F. Pereira, F.​​​‌Fabrizio Piras, F.​Federica Piras, S.​‌Sara Poletti, A.​​ E.Andrew E Reineberg​​​‌, G.Gloria Roberts​, R.Rafael Romero-García​‌, M. D.Matthew​​ D Sacchet, G.​​​‌ A.Giovanni A Salum​, A.-L.Anca-Larisa Sandu​‌, C. M.Carl​​ M Sellgren, E.​​​‌Eiji Shimizu, H.​ R.Harry R Smolker​‌, J. C.Jair​​ C Soares, J.​​​‌ D.J. Douglas Steele​, F.Frederike Stein​‌, D. J.Dan​​ J Stein, B.​​​‌Benjamin Straube, L.​Lea Teutenberg, F.​‌Florian Thomas-Odenthal, P.​​Paula Usemann, R.​​​‌Romain Valabregue, J.​Johanna Valencia-Echeverry, G.​‌Gerd Wagner, G.​​Gordon Waiter, M.​​​‌Martin Walter, H.​ C.Heather C Whalley​‌, M.-J.Mon-Ju Wu​​, T. T.Tony​​​‌ T Yang, C.​ A.Carlos A Zarate​‌, A.Andre Zugman​​, G. B.Giovana​​​‌ B Zunta-Soares, K.​Kees van Heeringen,​‌ S. J.Sanne J​​ H van Rooij,​​​‌ N.Nic van der​ Wee, S.Steven​‌ van der Werff,​​ P. M.Paul M​​​‌ Thompson, H. P.​Hilary P Blumberg,​‌ A.-L.Anne-Laura van Harmelen​​, M. E.Miguel​​​‌ E Rentería, N.​Neda Jahanshad, L.​‌ S.Laura S van​​ Velzen, E.Emmanuelle​​​‌ Le Bars, V.-G.​ O.Victor Ortiz-García de​‌ la Foz, K.​​Kees van Heeringen,​​​‌ S. J.Sanne J​ H van Rooij,​‌ N.Nic van der​​ Wee, S.Steven​​​‌ van der Werff,​ A.-L.Anne-Laura van Harmelen​‌ and L.Lianne Schmaal​​. Transdiagnostic alterations in​​​‌ white matter microstructure associated​ with suicidal thoughts and​‌ behaviours in the ENIGMA​​ Suicidal Thoughts and Behaviours​​​‌ consortium.Translational Psychiatry​1512025,​‌ 429HALDOI
• 86​​ articleD.Dang Wei​​​‌, A.Anna Freydenzon​, O.Octave Guinebretiere​‌, K.Karim Zaidi​​, F.Fen Yang​​​‌, W.Weimin Ye​, N.Niklas Hammar​‌, K.Karin Modig​​, N. R.Naomi​​​‌ R Wray, M.​Maria Feychting, N.​‌Nadine Hamieh, B.​​Bruno Ventelou, B.​​​‌Beranger Lekens, L.​Laurene Gantzer, S.​‌Stanley Durrleman, A.​​Allan Mcrae, B.​​​‌Baptiste Couvy-Duchesne, F.​Fang Fang and T.​‌Thomas Nedelec. Ten​​ years preceding a diagnosis​​​‌ of neurodegenerative disease in​ Europe and Australia: medication​‌ use, health conditions, and​​ biomarkers associated with Alzheimer's​​​‌ disease, Parkinson's disease, and​ amyotrophic lateral sclerosis.​‌EBioMedicine113March 2025​​, 105585HALDOI​​​‌back to text
• 87​ articleX.Xin Yue​‌, Q.Qing Zhao​​, X.Xiaoling Liu​​​‌, J.Jianqiang Li​, J.Jing Bai​‌, C.Changwei Song​​, S.Suqin Liu​​​‌, R.Rodrigo Moreno​, Z.Zhikai Yang​‌, S. E.Stefano​​ E Romero, G.​​​‌Gabriel Jimenez and G.​Guanghui Fu. Morphology-enhanced​‌ CAM-guided SAM for weakly​​ supervised breast lesion segmentation​​​‌.Biomedical Signal Processing​ and Control116January​‌ 2026, 109509HAL​​DOI
• 88 articleC.​​Chenxi Zhao, J.​​​‌Jianqiang Li, Q.‌Qing Zhao, J.‌​‌Jing Bai, S.​​Susana Boluda, B.​​​‌Benoit Delatour, L.‌Lev Stimmer, D.‌​‌Daniel Racoceanu, G.​​Gabriel Jimenez and G.​​​‌Guanghui Fu. ADNP-15:‌ An Open-Source Histopathological Dataset‌​‌ for Neuritic Plaque Segmentation​​ in Human Brain Whole​​​‌ Slide Images with Frequency‌ Domain Image Enhancement for‌​‌ Stain Normalization.Innovation​​ and Research in BioMedical​​​‌ engineering466December‌ 2025, 100913HAL‌​‌DOIback to text​​back to text