La Recherche à l'ULB| Données Personnelles | Thématiques de recherche | Charges de cours | Publications |
DEBEIR Olivier
Unités
Responsable d'Unité : Oui
LISA (Laboratory of Image Synthesis and Analysis) brings together expertise in image processing and analysis, pattern recognition, image synthesis and virtual reality. Its LISA-IA unit focuses on the fields of image analysis and pattern recognition and develops new methods for 2D and 3D object segmentation, recognition or tracking, multi-modal image registration, as well as machine and deep learning methods for signal and image processing. In the latter context, research is being carried out on the ability to deal with imperfect (weak or noisy) annotations and on methods of evaluating algorithms in such situations where the ground truth is not available. Developed algorithms are related to biomedical and industrial applications. Following a problem-centered approach, the unit tackles all hardware and software aspects of the chain in multidisciplinary teams (MDs, biologists, engineers, computer scientists, mathematicians, as well as art historians and archaeologists) over multi-institutional collaborations to deliver functional applications. The research is funded both by institutional/public funds and industry collaborations. LISA's achievements include one patent, several highly cited biomedical papers, implementation of acquisition and thermoregulation devices for live cell imaging, multi-media event organization and international cultural heritage projects.
Digital Image Analysis in Pathology
DIAPath est une unité de recherche transdisciplinaire et interfacultaire (Facultés de Médecine et École polytechnique de Bruxelles) intégrée au « Center for Microscopy and Molecular Imaging » (CMMI, Biopark de Gosselies). Cette unité est le fruit d’une collaboration de longue date entre le Service d’Anatomie Pathologique de l’Hôpital Erasme et le Laboratoire de l’Image : Synthèse et Analyse (LISA, Ecole polytechnique, ULB). Grâce à cette collaboration, DIAPath développe une approche intégrée de pathologie computationnelle pour la caractérisation, la validation et le monitoring de biomarqueurs histopathologiques au sein de tissus animaux et humains. L’approche développée par DIAPath fait appel aux techniques histologiques, d'immunohistochimie (IHC) et d'hybridation in situ chromogénique (CISH). En outre, l'unité a développé l'imagerie sur lame entière (Whole Slide Imaging) pour la caractérisation objective et quantitative de biomarqueurs au moyen de l'analyse d'image aidée de l’intelligence artificielle. Ces biomarqueurs peuvent être de nature morphologique ou concerner l'expression, la colocalisation ou la co-expression d'antigènes (ou d'autres molécules marquées), ainsi que leur distribution dans les échantillons histologiques. Une compétence en analyse de données vient compléter le dispositif. L'objectif général est d'extraire des informations utiles à la compréhension de processus pathologiques et de réponses aux traitements, ainsi que d'identifier et de valider de nouveaux biomarqueurs utiles à des fins diagnostiques, pronostiques et thérapeutiques. DIAPath poursuit son développement pour étendre ses techniques de marquage, d’imagerie et d’analyse à la fluorescence.
Projets
Development of 2D/3D vision and pattern recognition algorithm for the real time road traffic analysis and vehicle recognition (multi-modal image acquisition). Collaboration MACQ-electronic S.A.
Development of computer-based tools for the automatic tracking and the behavior analysis of cancerous cells evolving in in vitro 2D- or 3D-environment.
Development of 2D and 3D image analysis algorithm for the real time road traffic analysis and vehicle recognition. Collaboration MACQ-electronic S.A.
Partners: France, Belgique, Autriche, Mexique, Venezuela, Chili The Ipeca projects aims to create relation between European universities and universities of latin America, by phd and post doc exchanges around the common thematic and summer school organization.
Aide au diagnostic du mélanome par l'analyse d'image
Development of image analysis techniques for the melanoma early diagnosis aid
Symbol recognition, and in particular optical character recognition (OCR), can be consider as a mature domain. However when it comes to structured document, in particular schematics or blueprints, the complexity and the specificity of the document is increasing dramatically. The usual approach relies on line identification, symbol recognition and more generally OCR limited to portion of the document. Often ontologies, that are machine-compatible description of concepts, are defined to describe/analyse the document in a structured way. Ontologies are domain specific and require an important expert input. Blueprint project proposes to implement and train state of the art techniques of machine vision in order to provide the user with an interactive constantly learning system that will reinforce its prediction accuracy by a continuous human interaction. Moreover the chosen approach will enable to apply the same framework to dataset from diverse origin, in order to cope with the versatility of the blueprint digitization problem.
PICRIB-Platform for Imaging in Clinical Research in Brussels
Partners: Department of Electronics and Informatics (ETRO), VUB;Radiology department (RD), ULB - Hôpital Erasme (HE);Radiology department (RD), UCL-Cliniques Universitaires Saint-Luc (CUSL); Radiology Department (RD), VUB-UZBrussel;Laboratories of Image, Signal processing and Acoustics (LISA), LB;Department of Translational Research, Radiotherapy and Imaging (TRI), EORTC. Development of a clinical imaging platform in support of the clinical and research activities involving imaging, in particular for Brussels based academic research groups, the pharma and devices industry and SMEs.
Contribution to chinese character segmentation
Partner: Northwestern Polytechnical University
PROTHER-WAL : Proton Therapy Research in Wallonia
Image acquisition and processing for planning and monitoring proton therapy treatment (WP4) From macro (in vivo) to micro (histology) for preclinical animal model, involving image co-registration and quantitative analysis of tissue-based biomarkers Aims: analysis of treatment effects on tumor (microenvironment, healthy tissue, ...), validation of PET/IRM tracers
Development of 2D/3D vision and pattern recognition algorithm for the real time road traffic analysis and vehicle recognition (multi-modal image acquisition). Collaboration MACQ-electronic S.A.
Fast finger print matching using GPU
Development of GPU (Graphical Processing Unit) based algorithm for a fast finger minutiae identification. Collaboration ZETES Industries S.A.
Towards the next generation of smart and visual multi-modal sensor networks
Partner: VUB
Whole slide imaging and analysis in digital pathology
Tissue-based biomarker characterization from whole slide image analysis using machine and deep learning and image registration. This also includes the development of methods able to deal with imperfect (weak or noisy) annotations and methods of evaluating algorithms in such situations where the ground truth is not available This research is carried out in close collaboration with the Pathology Department of the Erasme hospital and the DIAPath pole (https://www.cmmi.be/?page_id=12) of the Center for Microscopy and Molecular Imaging (CMMI, Biopark of Gosselies, ULB).
Application of assemblies of weakened classifiers to remote sensed image segmentation, in particular using exogeneous data. Partner : IGEAT (ULB).
MARCH: Multilevel Analysis by revisiting Classifier Hierarchy
Study of multi-classifier-based systems applied to image processing and industrial applications.
echOpen - EICHO : Everywhere Imagery Care with Handheld and Open
echOpen is an open and collaborative project, led by a multidisciplinary community of experts and senior professionals joined by makers and Fablabs members with the goal of developing the very first low-cost and open source ultrasound probe connecting to any smartphone device. This probe is the stethoscope of our modern mes. This initiative allows radical transformation of clinics by enabling universal access diagnosis orientation in hospitals, community medicine and medically undeserved areas, in southern and northern countries. With the support of a wide academic ecosystem, including APHP and many research centers and universities, institutional foundations, and a large multidisciplinary open community, the project headquartered at Hotel Dieu hospital in Paris, has succeeded in September 2017 in developing a functional prototype displaying the first Open Source medical quality ultrasound image. This was released with a full open source documentation. echOpen is now heading to industrialization.
Star Image Recognition and Its Application in Guidance for Spacecrafts
Partner: Northwestern Polytechnical University
Assessing Intellectual Property Relevant Similarities In Images Through Algorithmic Decision Systems
The project aims at defining whether and how algorithmic technologies could be used to help in assessing, intellectual property (IP) relevant similarities, with a focus on image recognition technologies. Such, algorithmic decision systems (ADS) are currently being developed and used by private companies for, the purposes of IP enforcement (monitoring infringing goods online, filtering out content) and, registration by IP Offices, outside of public scrutiny.
PROTHER-WAL : Proton Therapy Research in Wallonia
Image acquisition and processing for planning and monitoring proton therapy treatment (WP4) From macro (in vivo) to micro (histology) for preclinical animal model, involving image co-registration and quantitative analysis of tissue-based biomarkers Aims: analysis of treatment effects on tumor (microenvironment, healthy tissue, ...), validation of PET/IRM tracers
Patient-Derived Tumor Growth Modeling from Multiparametric Analysis of Combined Dynamic PET/MR Data
Developing and validating a mathematical tumor growth model driven by patient-specific simultaneous Positron Emission Tomography (PET) and Magnetic Resonance Imaging (MRI) data to improve surgery and radiotherapy planning, especially for two types of tumors: gliomas, and hepatocellular carcinoma (HCC). This research is carried out in close collaboration with the Department of nuclear medicine, Erasme Hospital.
Development of 2D and 3D image analysis algorithm for the real time road traffic analysis and vehicle recognition. Collaboration MACQ-electronic S.A.
Setup of computer based tools and 3D biological models (in-vitro and ex-vivo) for a realistic study of cancerous cell migration and an efficient identification of potentially active anti-motility molecules.
ARMURS: Automatic Recognition for Map Update by Remote Sensing
Development of new algorithm for the detection of change between data extracted from known cartographic database and new acquired remote sensed images. The project aims to build a software prototype to validate the development on actual data. The project is done in collaboration with three Ulb laboratories (Mlg, Igeat and Lisa) and the Royal Military School
MASAAI - Motor Ability and Sleep Analysis using AI and Depth Sensor
Partner: Minnt S.A. The project aims to develop all the algorithms and tools necessary for the transition to "artificial intelligence", based on 3D sensors (time of flight, structured light, stereo), preventing and analyzing the fall of elderly people, people in revalidation and people with psychogeriatric disorders in hospitals, rest homes , rest and care homes and service flats, as well as to develop a new actimetric algorithm, in particular for sleep analysis. This project will enable us to meet the scientific, technological and ergonomic challenges necessary to achieve the levels of reliability and usability required for the success of our eHealth service.
COllaborative Consortium for the early detection of LIver CANcer (COCLICAN)
Partners: Institut de Recherche pour le Développement (IRD), Marseille, France. Université Toulouse III-Paul Sabatier (UT3), Toulouse, France. Kauno Technologijos Universitetas (KUT), Kaunas, Lithuania. Assistance Publique – Hôpitaux de Paris (APHP), Paris, France. Fondation Mérieux (FMER), Lyon, France.
ARTURO: Augmented RealiTy Unilateral RehabilitatiOn
Scientific studies show that 50-80% of people who undergo amputation develop severe pain they locate in the amputated limb: it is called phantom pain. There is a type of therapy in which the patient uses a mirror to have the illusion of viewing the amputated limb when in reality he observes the reflection of the healthy limb. However, the use of a physical mirror makes the less realistic overall illusion. Based on this observation, we have developed an augmented reality system based on images provided by RGB+d camera. From these images, we construct a textured mesh that displays a realistic 3D image of the patient, in stereovision on a 3D TV or in an immersion helmet. We are making a real-time monitoring of the central position of the patient's body and we use it to apply a mirror effect. With this, the patient has the illusion of having two arms again. With an implementation designed for the graphics card (GPU), our application works in real time. The aim of the Arturo project is to create a spin-off company for that application.
In vitro cell motility analysis
Partners: Unibioscreen S.A. Development of new image analysis method for in-vitro high throughput drug screening using phase-contrast video microscopy.
ARIAC (Applications et Recherche pour une Intelligence Artificielle de Confiance)
As part of the dynamics of the Walloon AI programme of Digital Wallonia, its objective is to create IT tools based on trusted artificial intelligence that can offer a competitive advantage to the Walloon industrial sector.
