DESCRIPTION :
The MIMESIS team is at the forefront of innovation in the fields of scientific computing, machine learning, medical imaging, and control. We are an interdisciplinary team that collaborates closely with clinicians to develop new technologies that can help improve healthcare, in particular through computer-assisted interventions. Our core research activities take place in the biomechanical modeling of soft tissue and developing novel numerical methods for real-time computation. Our research results pave the way towards augmented reality during interventions, autonomous medical robotics, and creating digital twins for personalized operation planning.
MIMESIS and LN Robotics, a South Korean medical robotics company, previously collaborated on a research project on autonomous endovascular navigation. LN Robotics has developed AVIAR, a robotized intervention system that reduces clinician radiation exposure during cardiovascular procedures. Building on their successful initial collaboration, both partners are now launching a new project to integrate artificial intelligence capabilities into AVIAR. This project aims to enhance fluoroscopic guidance through real-time vessel visualization and enable automated navigation. Our team's state-of-the-art vessel localization method will be central to this effort, with this postdoctoral position focused on advancing it toward clinical application.
Mission confiée
In fluoroscopy-guided endovascular interventions, vessel visualization traditionally relies on contrast agent injection. However, these agents are toxic at high doses and cannot be continuously injected. To overcome this limitation, we recently propose a solution based on deformable 2D-3D registration. Our method uses a preoperative 3D model of the vessels and updates it in real-time to provide clinicians with continuous vessel visualization throughout the procedure. Accurate and continuous vessel tracking is fundamental for autonomous endovascular interventions, as it enables precise robotic navigation without repeated contrast injections while ensuring safety through constant visual feedback.
Extracting vessel information from fluoroscopic images presents significant challenges: poor image contrast, limited 2D perspective, and scarcity of clinical data for training and validating deep learning approaches. Our deformable 2D-3D registration method was developed with these limitations in mind, and demonstrated promising results. The postdoctoral researcher will build upon these initial successes to develop a robust vessel tracking method suitable for clinical translation. This ambitious project requires innovative solutions at the intersection of computer vision, deep learning, and medical imaging. Success will depend on close collaboration with engineers and clinical partners to ensure that the developed method meets the robustness, accuracy, and real-time performance requirements of interventional practice
Principales activités
The successful candidate will work on improving and validating our deep learning-based deformable 2D-3D vessel registration method. Key objectives include:
1. Enhancing robustness and accuracy of vessel tracking under challenging clinical conditions (motion artifacts, varying contrast, overlapping structures)
2. Integrating vessel tracking in the AVIAR robotic system
3. Developing experimental validation protocols
4. Extending the method to handle catheter tracking
5. Creating safety measures for reliable autonomous navigation
Code d'emploi : Analyste (autre) (h/f)
Domaine professionnel actuel : Conseillers et Analystes de Management
Temps partiel / Temps plein : Plein temps
Type de contrat : Contrat à durée indéterminée (CDI)
Compétences : Modélisation 3D, Intelligence Artificielle, Réseaux de Neurones Artificiels, Vision par Ordinateur, Réalité Augmentée, Répertoire de Données Cliniques, Python (Langage de Programmation), Machine Learning, Informatique Scientifique, Conception et Développement de Logiciel, Pytorch, Deep Learning, Sens de la Communication, Axé sur le Succès, Esprit d'Équipe, Motivation Personnelle, Recherche, Écoute Active, Anatomie, Imagerie Médicale, Cardiologie, Travaux Cliniques, Fluoroscopie, Soins de Santé, Gestion de l'Innovation, Analyse Numérique, Recherche Post-Doctorale, Visualisation, Conception et Réalisation en Robotique, Protocoles de Validation, Pose de Sonde, Radio-oncologie, Applications des Règles et Consignes de Sécurité
Courriel :
Stephane.Cotin@inria.fr
Téléphone :
0139635511
Type d'annonceur : Employeur direct