Project A — Predictive Insights

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Curabitur blandit tempus porttitor.

• 🎯 Goal: Lorem ipsum dolor sit amet.
• 🛠️ Stack: Python, FastAPI, Postgres.
• 📈 Outcome: +32% efficiency.

Project B — Collaborative Analytics

Aenean eu leo quam. Pellentesque ornare sem lacinia quam venenatis vestibulum.

Highlights

1. Realtime dashboards
2. Role-based access
3. Automated reporting

Project C — ML Platform

Vestibulum id ligula porta felis euismod semper. Donec sed odio dui.

// Pseudo-API
POST /api/train
{ "model":"xgb", "cv":5, "metrics":["auc","f1"] }

Docs: Lorem ipsum dolor sit amet…

Jean-Philippe Villemin

Postdoctoral Researcher in Bioinformatics

Who I Am

I’m a postdoctoral researcher in bioinformatics specializing in cancer genomics and digital pathology. After ten years as a bioinformatics engineer across a university hospital (CHU), the Institut de Génétique in Montpellier, and the Centre Léon Bérard in Lyon, I defended my PhD in 2021. I now focus on the genomics on the High-grade serous ovarian carcinoma using spatial Transcriptomics.I care about reproducible science and translating omics + WSI into clinically useful insights. Outside the lab, I like kittens, good beer, and running.

Project

Exploring the heterogeneous landscape of High-Grade Serous Ovarian Carcinoma with Spatial Transcriptomics.

High-grade serous ovarian carcinoma (HGSOC) is an aggressive malignancy marked by extreme cellular heterogeneity, rapid progression, and frequent relapse after standard therapy. Current management generally combines cytoreductive surgery with platinum/taxane chemotherapy and targeted agents such as PARP inhibitors, yet overall survival at five years remains below 50%.

A major unmet need is precise patient stratification: we still cannot reliably predict who will respond to which treatment, or who will develop resistance. This limitation reflects the coexistence of multiple tumor clones and a complex, spatially organized microenvironment that is not fully understood.

In this project, we apply spatial transcriptomics to a prospective HGSOC cohort with numerous samples collected from multiple anatomical sites (adnexa, omentum, peritoneum). This approach allows us to map, in situ, the clonal architecture of the tumor, the local stromal landscape, and the site-specific transcriptional programs that may drive aggressiveness. We will also discuss analytic strategies to build and interpret this spatial atlas, including methods to annotate spatial transcriptomic data, to decipher intra- and inter-tumoral heterogeneity, and to address the computational challenges posed by the scale and complexity of these high-dimensional datasets.

Publications

Addressing multiple facets of ligand-receptor network inference including single-cell proteomics. Villemin JP, Pierre Giroux, Morgan Maillard, Pierre-Emmanuel Colombo, Christel Larbouret, Jacques Colinge. Nucleic Acid Research (2025) - Under Review. https://doi.org/10.1101/2025.10.05.680519

Differential predictive value of resident memory CD8+T cell subpopulations in non-small-cell lung cancer patients treated by immunotherapy. Léa Paolini,Thi Tran,Stéphanie Corgnac, Villemin JP, …,Eric Tartour. Journal of Immunotherapy of Cancer (2024)

Farnesyltransferase inhibition overcomes oncogene-addicted non-small cell lung cancer adaptive resistance to targeted therapies. Sarah Figarol, Célia Delahaye, Rémi Gence, Aurélia Doussine, Juan Pablo Cerapio, Mathylda Brachais, Claudine Tardy, Nicolas Béry, Raghda Asslan, Jacques Colinge, Jean-Philippe Villemin, Antonio Maraver, Irene Ferrer, Luis Paz-Ares, Linda Kessler, Francis Burrows, Isabelle Lajoie-Mazenc, Vincent Dongay, Clara Morin, Amélie Florent, Sandra Pagano, Estelle Taranchon-Clermont, Anne Casanova, Anne Pradines, Julien Mazieres, Gilles Favre & Olivier Calvayrac. Nat Commun (2024)

Inferring ligand-receptor cellular networks from bulk and spatial transcriptomic datasets with BulkSignalR. Villemin JP, Laia Bassaganyas, Didier Pourquier, Florence Boissiere, Simon Cabello-Aguilar, Evelyne Crapez, Rita Tanos, Emmanuel Cornillot, Andrei Turtoi, Jacques Colinge. Nucleic Acid Research (2023).

A cell-to-patient machine learning transfer approach uncovers novel basal-like breast cancer prognostic markers amongst alternative splice variants.

Villemin JP., Lorenzi, C., Cabrillac, MS. et al.

BMC Biology (2021). https://doi.org/10.1186/s12915-021-01002-7

Endothelial, epithelial, and fibroblast cells exhibit specific splicing programs independently of their tissue of origin. JP Villemin*, Mallinjoud P*, Mortada H*, Polay Espinoza M, Desmet FO, Samaan S, Chautard E, Tranchevent LC, Auboeuf D.

Genome Res. 2014 doi: 10.1101/gr.162933.113.

Charles Berger

PhD Student in Medical Imaging and AI

Who I Am

I am a Bioengineer from the École Polytechnique Fédérale de Lausanne (EPFL), where I completed a Master’s degree in Biophotonics and Bioimaging. Throughout my studies, I completed several internships ranging from automatic bone segmentation using active parametric contours on CT images to ECG signal reconstruction (BIG Lab, EPFL), as well as quality assessment research (SIMED, HUG).

These experiences strengthened my desire to build a career at the intersection of medical imaging, artificial intelligence, and clinical translation. I am particularly motivated by the collaborative and multidisciplinary nature of this field, which brings together expertise in medicine, bioinformatics, and image processing—an essential combination for successfully integrating AI into clinical practice.

As such, I am pursuing a PhD under the direction of Prof. Nougaret and co-supervise with Dr. Subsol (CNRS researcher at LIRMM). My research focuses on unravelling tumor biology in high-grade serous ovarian cancer through CT/MRI-based radiomics features, with the goal of developing virtual biopsy applications. This PhD is co-funded by the “IA pour la santé” (ANITI) program and INSERM.

What I Care About

My current work is centered on improving the reproducibility, explainability, standardization, and transparency of AI pipelines—both in deep learning and machine learning—as well as in radiomics feature extraction.

To achieve this, I emphasize robust environment configuration, version control, and the use of modern tools and frameworks, such as UV (package management), PyTorch Lightning, Hydra, MONAI, and Hugging Face Datasets.

Looking ahead, I am particularly interested in advancing explainability methods, which I believe are a critical and necessary step for the safe and trustworthy application of AI in clinical settings.

My Hobbies

Outside of research, I enjoy playing video games, watching anime, playing guitar (hoping to get back into it soon!), indoor climbing, weight training, reading, and building miniature models. I’m also a big animal lover, especially when it comes to cats.

My Project

Unravelling tumor biology in ovarian cancer with CT radiomics based features : the era of virtual biopsies (Project OCAS)

This project targets high-grade serous ovarian carcinoma (HGSOC), known for its advanced metastatic presentation and variable response to neoadjuvant chemotherapy (NACT) followed by delayed primary surgery (DPS). Despite the current treatment regimen, approximately 39% of patients do not benefit, highlighting the need for improved therapeutic approaches.

We acknowledge the complexity of HGSOC, from visible metastatic tumor volumes to microscopic tumor-immune environments and genomic diversity. Our goal is to enhance patient care by identifying subpopulations before treatment, such as those unlikely to respond to NACT, who might benefit more from immediate surgery

Historically, predictive studies have primarily relied on single data streams, such as clinical features or CT imaging. However, the potential of integrative, multi-omic models—which require large and well-annotated datasets—remains largely underexplored. Recent advances in Artificial Intelligence for precision oncology are increasingly centered on multimodal and multitask approaches. These models not only generate multiple outputs for distinct but correlated tasks but also integrate heterogeneous data types (e.g., text, tabular data, medical images). Such integration enables the combination of complementary information across biological scales—for instance, linking microscopic data (gene expression, histopathology) with macroscopic data (MRI scans, clinical reports)—to achieve a more holistic understanding of disease biology. Our project, OCAS, aims to advance this paradigm by leveraging two comprehensive datasets encompassing clinical data, chemotherapy response, CA-125 levels, circulating tumor DNA (ctDNA), and radiomic features at diagnosis. These multimodal data will be integrated into an ensemble machine-learning model designed to predict volumetric response to neoadjuvant chemotherapy (NACT), with a particular emphasis on the predictive power of radiomic features.

OCAS consists of three key components:

Development of an auto-segmentation system: Leveraging a large database of annotated radiological images and deep learning to accurately identify carcinomatous regions, enabling early tumor detection and personalized treatment strategies. This tackles thematics around model evaluation pipeline as well as metric selection for model evaluation and segmentation quality assessment.

Extraction of radiomic and pathomic data: Analyzing segmented images to derive radiomic and pathomic features, providing deeper insights into tumor characteristics and aiding in the discovery of precise biomarkers for better prognostics and treatment responses. This is where mutlimodality and multitask model architectures are deevlopped, especially for combining clinical data with radiomics features and also radiomics maps.

Identification of distinct clusters: Applying machine learning to pathomic and radiomic data to identify HGSOC subgroups with unique behaviors and susceptibilities, facilitating a precision medicine approach tailored to individual tumor characteristics (through Habitat Imaging method). By advancing our understanding and management strategies for HGSOC, OCAS aims to significantly improve treatment personalization and patient outcomes, representing a major step forward

Chad Estoup--Streiff

AI Engineer & PhD Student in Medical Imaging

Who I Am

I’m an AI and software engineer developing machine learning systems for medical imaging and multimodal biomedical data. My PhD focuses on integrating radiomics, histopathology, and genomic data to improve ovarian cancer diagnosis and treatment prediction. I’m passionate about building reliable and explainable AI models that bridge research and clinical practice.

Project

My current research aims to design deep learning architectures capable of processing complex medical data — including CT, MRI, and whole-slide images — within unified, multimodal frameworks. I work on reproducible training pipelines, feature harmonization, and explainability modules for clinical deployment.

Technical Skills

Expert in Python, PyTorch, and computer vision, I build AI models and infrastructure from data curation to deployment. My experience includes deep learning, radiomics, multimodal fusion, and high-performance computing with Docker, FastAPI, and MLflow. I prioritize clean, modular code and reproducible science.

Beyond Research

I enjoy all kind of science and technology topics or any meaningful projects. I also love sports such as climbing, hiking and wrestling.

Florent Tachenne

AI Engineer & PhD Student in Medical Imaging

Who I Am

I graduated from Phelma Grenoble INP engineering school, specializing in signal and image processing. My different internships led me to learn more about various deep learning applications such as segmentation, detection and inverse problems in medical imaging. Globally, my interest focuses on the use of mathematical tools to deal with medicine related problems in clinical settings. Apart from that, my life revolves around discovering new horizons through sports (soccer, running, etc), photography and cooking.

My work in the lab

I am currently pursuing a PhD (CIFRE with Siemens Healthineers) in the PINKcc lab, where my research focuses on using MRI data to predict radiotherapy response in locally advanced pancreatic ductal adenocarcinoma. My first project is to combine radiomics and dosimetry data from MR-guided radiotherapy to identify predictive biomarkers. Second, and most importantly, my thesis aims to translate new hypoxia-related biomarkers into clinical practice by implementing quantitative MR sequences—such as CEST and BOLD-based imaging—on a 3T magnet. This involves addressing motion-related artifacts, improving reproducibility, and reducing acquisition time through image processing and Physics-Informed Neural Networks. Finally, I am also developing semi-supervised and uncertainty-based deep learning methods to enhance pancreatic cancer segmentation and improve follow-up accuracy.

Margaux Verdier

Postdoctoral

Who I Am

I have been a postdoctoral researcher since September 2023 in the PINKcc lab at the Montpellier Cancer Research Institute (IRCM, INSERM U1194).

I obtained my PhD in 2022 in computer science applied to medical imaging at the Institute of Functional Human Imaging (I2FH) in Montpellier, in partnership with Siemens Healthineers.

My doctoral work focused on identifying MRI biomarkers predictive of tumor growth in diffuse low-grade gliomas, as well as developing automated tools for segmentation and transformation prediction.

My expertise spans MRI, clinical and preclinical imaging, computational imaging, and imaging data organization and preprocessing, complemented by strong skills in radiomics analysis, machine and deep learning, and programming in Python and MATLAB.

My Research Project

As a postdoctoral researcher, I specialize in ultra-high-field preclinical imaging (9.4T MRI, BNIF – University of Montpellier) for the advanced quantitative characterization of complex tumor tissues, and in the development of robust radiomic pipelines designed to create virtual biopsies for high-grade serous ovarian cancer and pancreatic cancer.

Both cancers share a major challenge: an extremely poor prognosis driven by late diagnosis and an incomplete understanding of their biological heterogeneity. Conventional biopsies sample only a small fraction of the tumor and cannot be repeated routinely, whereas non-invasive imaging coupled with radiomics offers a powerful alternative by extracting quantitative, reproducible descriptors from medical images.

My current project aims to develop a multiscale radiomics model of tumors by integrating three imaging levels: clinical MRI (1.5T/3T), preclinical ultra-high-field MRI (9.4T), and digitized histology. Leveraging ex vivo 9.4T MRI provides exceptional spatial resolution and fully controlled acquisition conditions, enabling precise characterization of tumor microstructure and intratumoral heterogeneity.

By comparing radiomic features across in vivo imaging, ex vivo imaging, and histology, this work aims to identify robust multiscale imaging signatures linked to histological and potentially molecular profiles. Ultimately, these imaging biomarkers could enhance early detection, support treatment-response prediction, and contribute to non-invasive patient stratification, bridging the gap between imaging and tumor biology within a precision-medicine framework.

My Publications

• Verdier M, Deverdun J, Le Bars E, et al. Evaluation of a nnU-Net type automated clinical volumetric tumor segmentation tool for diffuse low-grade glioma follow-up. J Neuroradiol. 2023 Jun 10:S0150-9861(23)00213-4. doi: 10.1016/j.neurad.2023.05.008
• Dos Santos T, Deverdun J, Chaptal T, Darlix A, Duffau H, Van Dokkum LEH, Coget A, Carrière M, Denis E, Verdier M, Menjot de Champfleur N, Le Bars E. Diffuse low-grade glioma: What is the optimal linear measure to assess tumor growth? Neurooncol Adv. 2024 Mar 27;6(1):vdae044. doi: 10.1093/noajnl/vdae044.

Marion Tardieu

CR ICM in physics

Who I Am

I am a physicist specializing in magnetic resonance imaging (MRI), with a focus on developing innovative biomarkers for tumor characterization. I joined IRCM (INSERM, UMR 1194) in 2018 as a postdoctoral researcher, where I worked on advancing MRI-based biomarkers for tumor assessment. In 2023, I secured a permanent position as a researcher at the institute (ICM affiliation).

I obtained my PhD in 2014 at IR4M (UMR 8081, Université Paris-Sud), where my research focused on MR elastography. My thesis aimed to develop, characterize, and optimize a system for inducing mechanical waves in deep organs to measure their mechanical properties. I also contributed to an original motion correction scheme for MR elastography, improving the accuracy and reliability of tissue stiffness measurements.

I then joined the Inflammation Research Center (CRI INSERM, UMR 1149) in 2015 as part of the EU-funded FORCE project (Horizon 2020). My role focused on measuring the apparent elasticity of liver tumors at varying levels of pre-compression using MRI elastography, working with a mouse model.

Projects

My work focuses on integrating innovative MRI biomarkers, developed in collaboration with the BioNanoImaging Foundry platform (BNIF, University of Montpellier), where I explore tumor physiology and treatment responses in depth.

We have developed multiparametric MRI biomarkers for comprehensive tumor characterization, including:

• Tissue elasticity via MR elastography,
• Tumor hypoxia using the TOLD-MRI method,
• Intratumoral acidity (pH) through CEST-MRI (co-supervision of a PhD thesis),
• Spatial heterogeneity via radiomics.

Each parameter is histologically validated (IHC, SHG, AFM) and acquired in a single session, enabling the generation of virtual biopsy maps using artificial intelligence. This approach opens new perspectives for precision medicine, offering non-invasive, dynamic, and predictive assessments of tumors and their response to therapies.

Publications

• Tardieu, M.; Lakhman, Y.; Khellaf, L.; Cardoso, M.; Sgarbura, O.; Colombo, P.-E.; Crispin-Ortuzar, M.; Sala, E.; Goze-Bac, C.; Nougaret, S. Assessing Histology Structures by Ex Vivo MR Microscopy and Exploring the Link Between MRM-Derived Radiomic Features and Histopathology in Ovarian Cancer. Frontiers in Oncology, 2022, 11.
• Tardieu, M.; Salameh, N.; Souris, L.; Rousseau, D.; Jourdain, L.; Skeif, H.; Prévot, F.; de Rochefort, L.; Ducreux, D.; Louis, B.; Garteiser, P.; Sinkus, R.; Darrasse, L.; Poirier-Quinot, M.; Maître, X. Magnetic Resonance Elastography with Guided Pressure Waves. NMR in Biomedicine, 2022, 35(7), e4701. https://doi.org/10.1002/nbm.4701
• Michalet, M.; Valenzuela, G.; Debuire, P.; Riou, O.; Azria, D.; Nougaret, S.; Tardieu, M. Robustness of Radiomics Features on 0.35 T Magnetic Resonance Imaging for Magnetic Resonance-Guided Radiotherapy. Physics and Imaging in Radiation Oncology, 2024, 31, 100613. https://doi.org/10.1016/j.phro.2024.100613
• Pagé, G.; Tardieu, M.; Gennisson, J.-L.; Besret, L.; Garteiser, P.; Van Beers, B.E. Tumor Solid Stress: Assessment with MR Elastography under Compression of Patient-Derived Hepatocellular Carcinomas and Cholangiocarcinomas Xenografted in Mice. Cancers, 2021, 13(8), 1891. https://doi.org/10.3390/cancers13081891
• Pagé, G.; Tardieu, M.; Besret, L.; Blot, L.; Lopes, J.; Sinkus, R.; Beers, B. E. V.; Garteiser, P. Assessing Tumor Mechanics by MR Elastography at Different Strain Levels. Journal of Magnetic Resonance Imaging, 2019, 50(6), 1982–1989. https://doi.org/10.1002/jmri.26787

Gaelle Boustany

PH, ICM

Dr. Gaëlle Boustany is a radiologist at the Montpellier Cancer Center, specializing in genitourinary imaging with a particular focus on prostate cancer. Trained in general oncologic radiology, she contributes to multidisciplinary care across the department while advancing imaging-based assessment of prostate disease, including staging, treatment response, and post-therapy surveillance. Her clinical expertise supports the integration of advanced MRI protocols and structured reporting to improve diagnostic precision and patient outcomes.

Ana Rusnac

Research Fellow, Romania

Dr. Ana Rusnac is a radiologist with a dedicated interest in gynecologic cancer imaging, especially ovarian and uterine malignancies. She combines clinical practice with two protected research days per week within the PINKCC Lab, where she plays a key role in clinical data curation and ovarian cancer segmentation for multi-omics and AI-driven projects. Her work supports large-scale radiogenomic studies aimed at refining disease characterization and improving patient stratification.

Laura Haddad

Research Fellow, Lebanon

Dr. Laura Haddad is a radiologist at the Montpellier Cancer Center with a strong involvement in gynecologic oncologic imaging. Alongside her clinical activity, she devotes two days per week to research within the PINKCC Lab, contributing to clinical dataset assembly and advanced ovarian segmentation for precision-medicine initiatives. Her participation in multi-institutional projects helps bridge routine imaging with emerging biomarkers to support individualized treatment planning.

Raphael Tetreau

PH, ICM

Dr. Raphaël Tétreau is the Head of the Radiology Department at the Montpellier Cancer Center, specializing in musculoskeletal oncology. His clinical practice focuses on the imaging of primary bone and soft-tissue tumors, metastatic disease, and post-treatment surveillance. He leads departmental strategy in innovation, education, and multidisciplinary integration, and supports the development of advanced imaging techniques and data-driven approaches to improve cancer diagnosis and patient management.

Pierre-Emmanuel Colombo

MD-PhD

Pr. Pierre-Emmanuel Colombo is Professor of Surgical Oncology at the Montpellier Cancer Center (ICM) and an internationally recognized expert in the management of gynecologic cancers, particularly advanced ovarian cancer. His clinical practice focuses on cytoreductive surgery combined with hyperthermic intraperitoneal chemotherapy (HIPEC), with a strong commitment to optimizing patient selection, improving perioperative outcomes, and refining strategies for recurrent disease. His research program explores prognostic and predictive factors, integration of biomarkers, and evaluation of innovative therapeutic approaches in high-risk settings. Pr. Colombo works closely with the PINKCC Lab to develop imaging- and AI-based tools that support resectability prediction and personalized treatment pathways, bridging surgical expertise with translational research. Actively involved in national and international collaborative networks, he plays a leading role in advancing standards of care and training the next generation of specialists in gynecologic oncology.