Industrial Engineer in Computer Science (Ing./M.Sc.) from ISIB
About me
Digital Business Analyst, Ph.D. in Computer Science
Currently, I am Digital Business Analyst at Fujitsu (BE). Prior to that, I have been working as a Scientific Software Engineer at CISTIB/UPF (ES), R&D Engineer in Intelligent Systems at Inspiralia (ES), Consultant in Energy Efficiency at E²=MC (BE), Consultant in Software Implementation at Sofico (BE), and Business Analyst at FIS (BE).
Since January 2011, I am Ph.D. in Computer Science from Mines ParisTech (FR). My Ph.D. was done at Asclepios/INRIA Sophia-Antipolis (FR). My thesis mainly deals with the analysis of medical images and took place within the framework of the EU Marie Curie project 3D Anatomical Human (3DAH).
Before starting my Ph.D., I took courses in computer vision and robotics at ViCOROB/UdG (ES). In 2005, I graduated Industrial Engineer in Computer Science (Ing./M.Sc.) from ISIB (BE).
Acronyms
CISTIB
Computational Imaging and Simulation Technologies in Biomedicine (ES)
FIS
Fidelity National Information Services Inc. (US)
INRIA
Institut National de Recherche en Informatique et en Automatique (FR)
ISIB
Institut Supérieur Industriel de Bruxelles (BE)
UdG
Universitat de Girona (ES)
UPF
Universitat Pompeu Fabra (ES)
ViCOROB
Visió per Computador i Robòtica (ES)
Social networks
My professional experience
Artificial Intelligence
Financial Services
Fleet Management
Energy Efficiency
Image Processing
Medical Imaging
Where I work
Where I worked
Where I studied
Positions I held
Business Analyst
Consultant
R&D Engineer
Software Engineer
My diplomas
Ph.D.
Ing./M.Sc.
For whom I work (and worked)
Where I studied
My recommendations
It was my pleasure to first meet François back in 2007-2008, on the occasion of the EU project 3D Anatomical Human. We coincided in many project meetings and workshops through the following years, and even spend some time visiting him and his lab in May 2010, at INRIA Sophia-Antipolis. I would like to emphasize that François always showed a very collaborative and constructive attitude towards work and during our meetings. He was an easy going researcher, that helps communication being fluid and efficient. He likes professionalism and good work being done and excelled in his research. In particular, I could see some of his work related to medical imaging and I really enjoyed both his work and presenting skills. Despite we didn't coincide on our specific research topics at that time, I should say that I would be pleased to work with him or collaborate once again in the future. Moreover, François is a very nice and respectful person to deal with, and I highly recommend François to anyone looking for such a nice combination of professional and personal values.
We were in the same research group with Francois while working for our doctoral degrees at INRIA Sophia Antipolis. I had the chance to interact with him on a daily basis during this time.
I find Francois to be a brilliant individual who is strong in both technical and social skills, which is a rare combination. In his research, he devised advanced algorithms for analyzing medical images and implemented them in a professional fashion. He was involved in interdisciplinary EU projects where he had to work with other engineers on algorithmic problems and also communicate results to medical doctors. His unique talent set allowed him to handle both of these situations very easily.
Francois is also a great colleague who helps building a collegial work environment. He is helpful to others and very respectful. Overall, it was a pleasure to spend my doctoral years in the same environment with Francois.
François and I worked in the Asclepios research team in INRIA. Back then, François was a Ph.D. student and I was a research engineer. We were both working on biomedical image analysis with application to medicine and biology to assist prevention, diagnosis and therapy. François worked on a very challenging topic during his Ph.D. and did an amazing work. I had the chance to collaborate with him on several occasions. He demonstrated that he is very skilled, both in mathematics and programming (C++/Python). He was able to quickly assimilate, use and develop advanced scientific concepts. For short, he was an excellent Ph.D. student. He was also very appreciated by other team members for his interpersonal skills. François is indeed an excellent colleague, very friendly, open-minded, always smiling and super funny. I strongly recommend François both as an engineer and as a colleague, and I hope I'll have the opportunity to work with him again.
François and I are both working at FIS Belgium within the CBK team, I as Head of Development for EMEA Delivery Center and François as Business Analyst in charge of customer-specific software delivery management. We had and still have the opportunity to work together for some common projects and customers, whether for the analysis and documentation of new CBK implementations, or for the analysis, support and testing of new enhancements. It is a pleasure to work with François, who is not only always eager to help and put the extra effort to do what needs to be done (especially at time of new software releases), but also fun to work with.
It was a pleasure to work with Francois in Inspiralia within the framework of the European project APIFRESH. Francois did an excellent job as scientific leader of the project; taking care of challenging technical and scientific subjects. He not only achieved spectacular results but also had time to write first class publications.
Francois was always hard worker and compromised with his work. His deep technical knowledge, practical mind, combined with long expertise in a wide number of subjects, has appointed Francois as one of our best scientific leaders. His honesty and friendship are also values to be highlighted.
I would like to recommend Francois Chung . We have been working together in the EU project 3D Anatomical Human I was coordinating. Francois was a great scientific collaborator in this project. He has worked intensively with my PhD students, particularly Jerome Schmitt, on the modelling of the hip. The collaboration has led to several publications together. He was a great scientist. A pity that this interdisciplinary project stopped after 4 years...
I wish Francois all the best.
Regards
Nadia Magnenat Thalmann
Founder and Director of MIRALab
University of Geneva.
thalmann@miralab.ch
François Chung est une personne enthousiaste et dynamique. J'ai pu collaborer avec lui sur la mise en oeuvre de formations dans le domaine de l'énergie. Il comprend rapidement et est doté d'un esprit de synthèse,. C'est très agréable de travailler avec lui. Il met les personnes à l'aise et entre facilement en contact.
I was a colleague of Francois for just under 2 years in the technical department of Inspiralia in Madrid. During that time, Francois demonstrated exceptional technical skills in his area of soft computing and mathematical modelling (evidenced by the publication of the results of one of his projects in a respected scientific journal). He was also a colleague with strong social skills and the ability to get on with everyone. I have never written a recommendation for anyone else on Linkedin and don´t intend to but in the case of Francois he would be an asset to any organisation that was lucky enough to employ him.
I had the luck to work with François in the project MySPine at CISTIB, where he was a software engineer and I was a biomedical researcher. He was a fount of information about C++ and full of new ideas to try out. But, what I miss the most about François is his positive attitude towards his work and to know that he would do his utmost to help you with any problem you have.
I met François when I was doing my PhD in VicoRob group (University in Girona) as we worked in the same lab. He is a very nice person, easy to talk with, that helps to create a good working team. Despite we work in different topics, he was always ready to help the colleagues when needed. A good friend inside and outside the working environment.
Francois is an extremely disciplined and self motivated individual. His attention to detail and ability to create and maintain business relationships are his key strengths. Francois has developed extensive business acumen within the Fleet and Leasing industries, far beyond his years of experience. During our years of working together I found Francois to be a very dependable and trustworthy Consultant. He would be an asset to any organisation and his energetic approach and strong consulting skills would benefit any business project.
I met François during the Marie Curie European Project "3D Anatomical Human" (3DAH). We shared particular interest in medical image segmentation and anatomical modeling. As a result, we collaborated together on medical image processing approaches which were using simplex meshes and most importantly advanced prior knowledge. In particular, François developed an expertise in intensity-based prior knowledge.
The result of the collaborator was a shared publication and a series of personalized anatomical models of the lower limb built from MR images. Besides his strong experience in programming and medical image analysis, François is an amazing co-worker, and if I would have the chance again I will work with him without any hesitation,
Oh by the way he is also a great drummer : )
I had the pleasure to work alongside François for two years at Inspiralia in Madrid.
François is passionate about the field of programming, artificial intelligence and mathematical modelling. He brings a set of technical skills which allow him to deliver high quality work, and communicate it effectively.
On top of that, François possesses great social skills and speaks several foreign languages at native-like level, making him a superb candidate to integrate international and multidisciplinary teams.
I cannot but endorse him.
When Francois enters the door of your office, he might look like a regular applicant. Wake up, he is so much more than that! This guy can discuss his cutting-edge scientific project over lunch and be so enthusiastic about it that you immediately start working with him forgetting about your coffee. And God knows coffee is sacred. Meticulous and dedicated when it comes to work. Socially aware and friendly when it comes to beer. We shared a great deal of scientific insights, inspirational ideas and pizzas while working hard towards our PhDs. Get him in your team and enjoy!
We worked together as consultants for a software product. This product was and is continuously evolving, and as such the knowledge of the consultants is put to the test all the time. I know François as being always inquisitive and eager to learn so as to improve in his job. Also in conversation it was always amable.
I met François while working on my master thesis at the University of Girona. We were both members of the Underwater Vision and Robotics lab from the ViCOROB research group. It was a pleasure to work with François, who is proactive, result oriented, fun to work with and always ready to put all his energy and time to get the job done. This is evidenced by the fact that the VICOROB-UdG team, in which he developed a program for the real-time tracking of underwater objects, won the SAUC-E 2006 competition.
Working with Francois on a customer implementation project was a very positive experience. Francois was involved in the same implementation project as I was and we worked closely together on a requirements gathering phase to determine the best business process for our customer. Francois works very precise and has a healthy view on the approach for business process analysis. His eagerness to learn, analytical skills and linguistic diversity give Francois a valuable combination of abilities to work a complex and international environment.
Business consultancy and business development support within the Digital Business Solutions (DBS) department, which covers artificial intelligence (AI), blockchain, robotic process automation (RPA), data analytics and quantum-inspired computing.
Main tasks: Gathering and analyzing business requirements from the customer; Supporting technical analysis to translate requirements into technical specifications; Creating functional deliverables according to business requirements; Managing transition and handover to the customer.
2016-2018
Business Analyst
FIS, Belgium
Business analysis, project coordination and product implementation of digital platforms that handle payments, disputed transactions and transactional services between a company, its partners and its customers.
Main tasks: Meeting the client to discuss its business needs; Project coordination during the software implementation phase; Installation and configuration of the solution within the client’s environment; Product presentation, user training and post project support.
2015-2016
Consultant in Software Implementation
Sofico, Belgium
Implementation of Miles, which is a flexible end-to-end software for automotive finance, leasing, fleet and mobility management, after the analysis of the client’s requirements and their translation into a configuration solution.
Main tasks: Business consulting (e.g. functional analysis and reporting); Participation in the technical aspects of the implementation (e.g. batch processing, interface, security, data mapping, testing); Budget determination for configuration and migration; Participation in pre-sales and commercial activities.
2013-2015
Consultant in Energy Efficiency
E²=MC, Belgium
Implementation of energy efficiency policies aimed at improving the rational use of energy and the energy performance of buildings (e.g. improvement of the thermal insulation and both heating and lighting systems).
Main tasks: Project management (e.g. supervision of subcontracted work, writing and management of both technical and activity reports); Analysis of energy efficiency-related works (e.g. using technical, energetic and economic criteria so as to determine their relevance and rank them according to their energy efficiency).
2012-2013
R&D Engineer
Inspiralia/Pera, Spain
Design, proposal and implementation of industrial projects based on intelligent systems (e.g. pattern recognition, data analysis, image processing and artificial intelligence) for the benefit of SMEs in European consortiums.
Main tasks: Participation in the data acquisition; Implementation of state-of-the-art algorithms; Prototype creation based on Matlab and C++ in a Windows environment; Supervision of subcontracted work; Writing of both technical reports and scientific publications.
2011-2012
Scientific Software Engineer
CISTIB, UPF, Spain
Participation in the development of GIMIAS, which is a workflow-oriented software framework whose objective is the fast prototyping of applications for solving advanced biomedical image computing and simulation problems.
Main tasks: Programming within the team using SVN; C++ plugin development in a Windows environment; Use of open source libraries (e.g. VTK, ITK, wxWidgets); Software optimization to speed up computation time.
2007-2010
Ph.D. Candidate
Asclepios, INRIA Sophia-Antipolis, France
Early Stage Researcher (ESR) in the EU Marie Curie project 3D Anatomical Human (3DAH). Main research topics include liver segmentation in CT images, bone segmentation in MR images and unsupervised clustering.
Main tasks: Programming within the team using CVS and SVN; State-of-the-art algorithm prototyping in Matlab; Software development in a Linux environment; Use of open source libraries (e.g. VTK, ITK, STL); Bash scripting; Code parallelization.
2006
Team Member
ViCOROB, UdG, Spain
Participation in the SAUC-E robotics competition within the VICOROB-UdG team, which won the 2006 edition in London. Development of a program based on image processing for the real-time tracking of underwater objects.
Main tasks: Real-time tracking of underwater objects using image processing; Programming in C in an embedded environment; Algorithm optimization to speed up computation time so as to reach real-time image processing (25 frames per second).
2005
Intern
EASI, Belgium
Implementation of a centralized supervision system, which consists in detecting failures from various infrastructures on the network (e.g. servers, printers and workstations) and in generating alarms in a software interface.
Main tasks: Installation and deployment of a supervision system in a multi-platform environment (e.g. Windows, Linux, Mac OS X, OS/400); Plugin development using both Shell and Perl scripts; Creation of a multi-user mode based on LDAP authentication.
Education
2007-2010
Mines ParisTech
École des Mines de Paris, France
Ph.D. in Computer Science (Image Processing)
2005-2007
UdG
Universitat de Girona, Spain
Ph.D. courses in Computer Vision and Robotics
2001-2005
ISIB
Institut Supérieur Industriel de Bruxelles, Belgium
Industrial Engineer in Computer Science (Ing.) Master in Industrial Engineer Sciences (M.Sc.)
Languages
French
English
Spanish
Dutch
Catalan
Technical skills
C++, Java, Matlab, Tcl/Tk, GDB, Valgrind
LaTeX, Microsoft Office, Open Office
Linux, Microsoft Windows
XHTML, PHP, CSS5, Javascript
MySQL, Oracle, DB2
Gimp, Inkscape, Jira
Acronyms
CSS
Cascading Style Sheets
CT
Computed Tomography
CVS
Concurrent Versions System
GDB
The GNU Debugger
GIMIAS
Graphical Interface for Medical Image Analysis and Simulation
ITK
The Insight Segmentation and Registration Toolkit
LDAP
Lightweight Directory Access Protocol
MR
Magnetic Resonance
PHP
Hypertext Preprocessor
SAUC-E
Student Autonomous Underwater Challenge – Europe
@misc{mines-2018-magazine-article,
title = {Une politique énergétique belge à multi-niveaux},
author = {François Chung},
year = {2018},
date = {2018-11-01},
abstract = {Au XIXe siècle, le développement économique de la Belgique fut propulsé par le secteur énergétique grâce aux charbonnages wallons. En effet, dans la 1ère moitié du XIXe siècle, la Belgique était un des seuls pays, avec l’Angleterre, où le charbonnage était à ce point développé. À un tel point que des mines de charbon belges furent les premières entreprises industrielles à entrer en bourse à Paris. Cette époque d’abondance énergétique contraste avec la situation telle que nous la connaissons aujourd’hui : la Belgique dépend désormais de l’étranger pour ses besoins énergétiques, que ce soit pour le pétrole, le gaz naturel ou le charbon lui-même, qui représentent tous trois un peu plus de 70% de sa consommation d’énergie primaire en 2014.},
howpublished = {Mines Revue des Ingénieurs, 499, pp. 27-29},
keywords = {efficacité énergétique, isolation thermique, marchés de l’énergie, système de chauffage},
pubstate = {published},
tppubtype = {misc}
}
Au XIXe siècle, le développement économique de la Belgique fut propulsé par le secteur énergétique grâce aux charbonnages wallons. En effet, dans la 1ère moitié du XIXe siècle, la Belgique était un des seuls pays, avec l’Angleterre, où le charbonnage était à ce point développé. À un tel point que des mines de charbon belges furent les premières entreprises industrielles à entrer en bourse à Paris. Cette époque d’abondance énergétique contraste avec la situation telle que nous la connaissons aujourd’hui : la Belgique dépend désormais de l’étranger pour ses besoins énergétiques, que ce soit pour le pétrole, le gaz naturel ou le charbon lui-même, qui représentent tous trois un peu plus de 70% de sa consommation d’énergie primaire en 2014.
@book{eue-2017-book,
title = {Multi-focal Image Segmentation, Classification and Authentication: A General Framework applied on Microscope Pollen Images},
author = {François Chung and Tomás Rodríguez},
isbn = {978-3841677907},
year = {2017},
date = {2017-02-20},
publisher = {Éditions universitaires européennes (EUE)},
address = {Saarbrücken},
abstract = {In this book, we propose a general framework for multi-focal image segmentation, classification and authentication, the methodology being demonstrated on microscope pollen images. The framework is meant to be generic and based on a brute force-like approach aimed to be efficient not only on any kind, and any number, of pollen images (regardless of the pollen type), but also on any kind of multi-focal images. All stages of the framework are designed to be used in an automatic fashion. First, the optimal focus is selected using the absolute gradient method. Then, pollen grains are extracted from microscope images, followed by the automatic segmentation of their exine. A coarse-to-fine approach ensures a smooth and accurate segmentation of both structures. Finally, feature extraction and selection are performed on pollen grains using a generalized approach and the pollen classification is tested with four classifiers: Weighted Neighbor Distance, Neural Network, Decision Tree and Random Forest. The latter method, which has shown the best and more robust classification accuracy results (above 97% for any number of pollen types), is finally used for a final authentication stage.},
keywords = {active contour model, image processing, microscope, optimal focus, pollen},
pubstate = {published},
tppubtype = {book}
}
In this book, we propose a general framework for multi-focal image segmentation, classification and authentication, the methodology being demonstrated on microscope pollen images. The framework is meant to be generic and based on a brute force-like approach aimed to be efficient not only on any kind, and any number, of pollen images (regardless of the pollen type), but also on any kind of multi-focal images. All stages of the framework are designed to be used in an automatic fashion. First, the optimal focus is selected using the absolute gradient method. Then, pollen grains are extracted from microscope images, followed by the automatic segmentation of their exine. A coarse-to-fine approach ensures a smooth and accurate segmentation of both structures. Finally, feature extraction and selection are performed on pollen grains using a generalized approach and the pollen classification is tested with four classifiers: Weighted Neighbor Distance, Neural Network, Decision Tree and Random Forest. The latter method, which has shown the best and more robust classification accuracy results (above 97% for any number of pollen types), is finally used for a final authentication stage.
@article{micron-2015-academic-journal-article,
title = {Automated pollen identification using microscopic imaging and texture analysis},
author = {Víctor J Marcos and Rodrigo Nava and Gabriel Cristóbal and Rafael Redondo and Boris Escalante-Ramírez and Gloria Bueno and Óscar Déniz and Amelia González-Porto and Cristina Pardo and François Chung and Tomás Rodríguez},
doi = {10.1016/j.micron.2014.09.002},
issn = {0968-4328},
year = {2015},
date = {2015-01-01},
journal = {Micron},
volume = {68},
number = {0},
pages = {36-46},
abstract = {Pollen identification is required in different scenarios such as prevention of allergic reactions, climate analysis or apiculture. However, it is a time-consuming task since experts are required to recognize each pollen grain through the microscope. In this study, we performed an exhaustive assessment on the utility of texture analysis for automated characterisation of pollen samples. A database composed of 1800 brightfield microscopy images of pollen grains from 15 different taxa was used for this purpose. A pattern recognition-based methodology was adopted to perform pollen classification. Four different methods were evaluated for texture feature extraction from the pollen image: Haralick's gray-level co-occurrence matrices (GLCM), log-Gabor filters (LGF), local binary patterns (LBP) and discrete Tchebichef moments (DTM). Fisher's discriminant analysis and k-nearest neighbour were subsequently applied to perform dimensionality reduction and multivariate classification, respectively. Our results reveal that LGF and DTM, which are based on the spectral properties of the image, outperformed GLCM and LBP in the proposed classification problem. Furthermore, we found that the combination of all the texture features resulted in the highest performance, yielding an accuracy of 95%. Therefore, thorough texture characterisation could be considered in further implementations of automatic pollen recognition systems based on image processing techniques.},
keywords = {discrete Tchebichef moments, gray-level co-occurrence matrix, Log-Gabor filters},
pubstate = {published},
tppubtype = {article}
}
Pollen identification is required in different scenarios such as prevention of allergic reactions, climate analysis or apiculture. However, it is a time-consuming task since experts are required to recognize each pollen grain through the microscope. In this study, we performed an exhaustive assessment on the utility of texture analysis for automated characterisation of pollen samples. A database composed of 1800 brightfield microscopy images of pollen grains from 15 different taxa was used for this purpose. A pattern recognition-based methodology was adopted to perform pollen classification. Four different methods were evaluated for texture feature extraction from the pollen image: Haralick's gray-level co-occurrence matrices (GLCM), log-Gabor filters (LGF), local binary patterns (LBP) and discrete Tchebichef moments (DTM). Fisher's discriminant analysis and k-nearest neighbour were subsequently applied to perform dimensionality reduction and multivariate classification, respectively. Our results reveal that LGF and DTM, which are based on the spectral properties of the image, outperformed GLCM and LBP in the proposed classification problem. Furthermore, we found that the combination of all the texture features resulted in the highest performance, yielding an accuracy of 95%. Therefore, thorough texture characterisation could be considered in further implementations of automatic pollen recognition systems based on image processing techniques.
@article{compag-2015-academic-journal-article,
title = {Pollen segmentation and feature evaluation for automatic classification in bright-field microscopy},
author = {Rafael Redondo and Gloria Bueno and François Chung and Rodrigo Nava and Víctor J Marcos and Gabriel Cristóbal and Tomás Rodríguez and Amelia González-Porto and Cristina Pardo and Óscar Déniz and Boris Escalante-Ramírez},
doi = {10.1016/j.compag.2014.09.020},
issn = {0168-1699},
year = {2015},
date = {2015-01-01},
journal = {Computers and Electronics in Agriculture (COMPAG)},
volume = {110},
number = {0},
pages = {56–69},
abstract = {Besides the well-established healthy properties of pollen, palynology and apiculture are of extreme importance to avoid hard and fast unbalances in our ecosystems. To support such disciplines, computer vision comes to alleviate tedious recognition tasks. In this paper, we present an applied study of the state of the art in pattern recognition techniques to describe, analyze, and classify pollen grains in an extensive dataset specifically collected (15 types, 120 samples/type). We also propose a novel contour-inner segmentation of grains, improving 50% of accuracy. In addition to published morphological, statistical, and textural descriptors, we introduce a new descriptor to measure the grain’s contour profile and a logGabor implementation not tested before for this purpose. We found a significant improvement for certain combinations of descriptors, providing an overall accuracy above 99%. Finally, some palynological features that are still difficult to be integrated in computer systems are discussed.},
keywords = {automatic classification, bright-field microscopy, feature extraction},
pubstate = {published},
tppubtype = {article}
}
Besides the well-established healthy properties of pollen, palynology and apiculture are of extreme importance to avoid hard and fast unbalances in our ecosystems. To support such disciplines, computer vision comes to alleviate tedious recognition tasks. In this paper, we present an applied study of the state of the art in pattern recognition techniques to describe, analyze, and classify pollen grains in an extensive dataset specifically collected (15 types, 120 samples/type). We also propose a novel contour-inner segmentation of grains, improving 50% of accuracy. In addition to published morphological, statistical, and textural descriptors, we introduce a new descriptor to measure the grain’s contour profile and a logGabor implementation not tested before for this purpose. We found a significant improvement for certain combinations of descriptors, providing an overall accuracy above 99%. Finally, some palynological features that are still difficult to be integrated in computer systems are discussed.
@article{cviu-2013-academic-journal-article,
title = {Regional appearance modeling based on the clustering of intensity profiles},
author = {François Chung and Hervé Delingette},
doi = {10.1016/j.cviu.2013.01.011},
issn = {1077-3142},
year = {2013},
date = {2013-06-01},
journal = {Computer Vision and Image Understanding (CVIU)},
volume = {117},
number = {6},
pages = {705-717},
abstract = {Model-based image segmentation is a popular approach for the segmentation of anatomical structures from medical images because it includes prior knowledge about the shape and appearance of structures of interest. This paper focuses on the formulation of a novel appearance prior that can cope with large variability between subjects, for instance due to the presence of pathologies. Instead of relying on Principal Component Analysis (PCA) such as in Statistical Appearance Models (SAMs), our approach relies on a multimodal intensity profile atlas from which a point may be assigned to several profile modes consisting of a mean profile and its covariance matrix. These profile modes are first estimated without any intra-subject registration through a boosted Expectation-Maximization (EM) classification based on spectral clustering. Then, they are projected on a reference mesh whose role is to store the appearance information in a common geometric representation. We show that this prior leads to better performance than the classical monomodal PCA approach while relying on fewer profile modes.},
keywords = {appearance modeling, model-based image segmentation, unsupervised clustering},
pubstate = {published},
tppubtype = {article}
}
Model-based image segmentation is a popular approach for the segmentation of anatomical structures from medical images because it includes prior knowledge about the shape and appearance of structures of interest. This paper focuses on the formulation of a novel appearance prior that can cope with large variability between subjects, for instance due to the presence of pathologies. Instead of relying on Principal Component Analysis (PCA) such as in Statistical Appearance Models (SAMs), our approach relies on a multimodal intensity profile atlas from which a point may be assigned to several profile modes consisting of a mean profile and its covariance matrix. These profile modes are first estimated without any intra-subject registration through a boosted Expectation-Maximization (EM) classification based on spectral clustering. Then, they are projected on a reference mesh whose role is to store the appearance information in a common geometric representation. We show that this prior leads to better performance than the classical monomodal PCA approach while relying on fewer profile modes.
@techreport{inspiralia-2013-technical-report,
title = {A general framework for multi-focal image classification and authentication: Application to microscope pollen images},
author = {François Chung and Tomás Rodríguez},
year = {2013},
date = {2013-01-01},
address = {Madrid},
institution = {Inspiralia,},
abstract = {In this article, we propose a general framework for multi-focal image classification and authentication, the methodology being demonstrated on microscope pollen images. The framework is meant to be generic and based on a brute force-like approach aimed to be efficient not only on any kind, and any number, of pollen images (regardless of the pollen type), but also on any kind of multi-focal images. All stages of the framework's pipeline are designed to be used in an automatic fashion. First, the optimal focus is selected using the absolute gradient method. Then, pollen grains are extracted using a coarse-to-fine approach involving both clustering and morphological techniques (coarse stage), and a snake-based segmentation (fine stage). Finally, features are extracted and selected using a generalized approach, and their classification is tested with four classifiers: Weighted Neighbor Distance, Neural Network, Decision Tree and Random Forest. The latter method, which has shown the best and more robust classification accuracy results (above 97% for any number of pollen types), is finally used for the authentication stage.},
keywords = {feature extraction, image processing, optimal focus, supervised learning},
pubstate = {published},
tppubtype = {techreport}
}
In this article, we propose a general framework for multi-focal image classification and authentication, the methodology being demonstrated on microscope pollen images. The framework is meant to be generic and based on a brute force-like approach aimed to be efficient not only on any kind, and any number, of pollen images (regardless of the pollen type), but also on any kind of multi-focal images. All stages of the framework's pipeline are designed to be used in an automatic fashion. First, the optimal focus is selected using the absolute gradient method. Then, pollen grains are extracted using a coarse-to-fine approach involving both clustering and morphological techniques (coarse stage), and a snake-based segmentation (fine stage). Finally, features are extracted and selected using a generalized approach, and their classification is tested with four classifiers: Weighted Neighbor Distance, Neural Network, Decision Tree and Random Forest. The latter method, which has shown the best and more robust classification accuracy results (above 97% for any number of pollen types), is finally used for the authentication stage.
@techreport{inspiralia-2012-technical-report,
title = {Automatic pollen grain and exine segmentation from microscope images},
author = {François Chung and Tomás Rodríguez},
year = {2012},
date = {2012-12-01},
address = {Madrid},
institution = {Inspiralia,},
abstract = {In this article, we propose an automatic method for the segmentation of pollen grains from microscope images, followed by the automatic segmentation of their exine. The objective of exine segmentation is to separate the pollen grain in two regions of interest: exine and inner part. A coarse-to-fine approach ensures a smooth and accurate segmentation of both structures. As a rough stage, grain segmentation is performed by a procedure involving clustering and morphological operations, while the exine is approximated by an iterative procedure consisting in consecutive cropping steps of the pollen grain. A snake-based segmentation is performed to refine the segmentation of both structures. Results have shown that our segmentation method is able to deal with different pollen types, as well as with different types of exine and inner part appearance. The proposed segmentation method aims to be generic and has been designed as one of the core steps of an automatic pollen classification framework.},
keywords = {exine, image processing, pollen grains, pollen identification, texture analysis},
pubstate = {published},
tppubtype = {techreport}
}
In this article, we propose an automatic method for the segmentation of pollen grains from microscope images, followed by the automatic segmentation of their exine. The objective of exine segmentation is to separate the pollen grain in two regions of interest: exine and inner part. A coarse-to-fine approach ensures a smooth and accurate segmentation of both structures. As a rough stage, grain segmentation is performed by a procedure involving clustering and morphological operations, while the exine is approximated by an iterative procedure consisting in consecutive cropping steps of the pollen grain. A snake-based segmentation is performed to refine the segmentation of both structures. Results have shown that our segmentation method is able to deal with different pollen types, as well as with different types of exine and inner part appearance. The proposed segmentation method aims to be generic and has been designed as one of the core steps of an automatic pollen classification framework.
@conference{vph-2012-conference-proceeding,
title = {A software framework for 3D reconstruction and fracture risk assessment of the proximal femur from dual-energy x-ray absorptiometry},
author = {Yves Martelli and Tristan Whitmarsh and Ludovic Humbert and François Chung and Luis M Río del Barquero and Silvana Di Gregorio and Luigi Carotenuto and Alejandro F Frangi},
year = {2012},
date = {2012-09-01},
booktitle = {VPH 2012: Virtual Physiological Human - Integrative approaches to computational biomedicine},
address = {London},
abstract = {In clinical routine, osteoporosis diagnosis is performed using the areal Bone Mineral Density (aBMD) computed from two-dimensional Dual-energy X-ray Absorptiometry (DXA) images. This measurement of the "projected density" gives a global marker of osteoporosis but is insufficient for an accurate quantification of non-uniform mineral deficiencies. Quantitative Computed Tomography (QCT) is one possible modality for retrieving shape and bone mineral density measurements in 3D, leading to a better characterization of the local bone quality and the fracture risk. However, due to high financial costs and a high radiation dose for the patient, this modality is rarely used in clinical routine for the follow-up of osteoporosis. In this work, we show how state-of-the-art techniques and methods for fracture risk assessment of the proximal femur using 3D analysis from single-view DXA images were brought to the research and clinical end-user, through the use of an existing open source framework for fast prototyping of clinical applications.},
keywords = {bone mineral density, DXA image, fracture risk, medical image processing},
pubstate = {published},
tppubtype = {conference}
}
In clinical routine, osteoporosis diagnosis is performed using the areal Bone Mineral Density (aBMD) computed from two-dimensional Dual-energy X-ray Absorptiometry (DXA) images. This measurement of the "projected density" gives a global marker of osteoporosis but is insufficient for an accurate quantification of non-uniform mineral deficiencies. Quantitative Computed Tomography (QCT) is one possible modality for retrieving shape and bone mineral density measurements in 3D, leading to a better characterization of the local bone quality and the fracture risk. However, due to high financial costs and a high radiation dose for the patient, this modality is rarely used in clinical routine for the follow-up of osteoporosis. In this work, we show how state-of-the-art techniques and methods for fracture risk assessment of the proximal femur using 3D analysis from single-view DXA images were brought to the research and clinical end-user, through the use of an existing open source framework for fast prototyping of clinical applications.
@misc{mines-2012-magazine-article,
title = {De l'imagerie médicale à la médecine du futur},
author = {François Chung},
year = {2012},
date = {2012-02-01},
journal = {Mines Revue des Ingénieurs},
number = {458},
abstract = {Depuis ses débuts, l'imagerie médicale a pour objectif de fournir aux radiologues des images médicales afin de les aider dans leur diagnostic. Avec l'avancée des techniques d'acquisition, les radiologues se retrouvent à analyser des images de plus en plus complexes et dans des quantités de plus en plus importantes. Du côté de la recherche, cela se traduit par une collaboration entre physique médicale, radiologie et imagerie médicale. Les physiciens ont pour objectif d'améliorer la qualité et la résolution des images médicales. Ces améliorations permettent d'aider les radiologues dans leur diagnostic et à la communauté de l'imagerie médicale de pouvoir extraire des informations plus précises. Cette collaboration permet non seulement d'avancer dans les sciences médicales (ex. étude de l'anatomie et physiologie), mais également dans les applications cliniques (ex. détection de maladies et planification de thérapie).},
howpublished = {Mines Revue des Ingénieurs, 458, pp. 53-56},
keywords = {imagerie médicale, simulation informatique, traitement du patient},
pubstate = {published},
tppubtype = {misc}
}
Depuis ses débuts, l'imagerie médicale a pour objectif de fournir aux radiologues des images médicales afin de les aider dans leur diagnostic. Avec l'avancée des techniques d'acquisition, les radiologues se retrouvent à analyser des images de plus en plus complexes et dans des quantités de plus en plus importantes. Du côté de la recherche, cela se traduit par une collaboration entre physique médicale, radiologie et imagerie médicale. Les physiciens ont pour objectif d'améliorer la qualité et la résolution des images médicales. Ces améliorations permettent d'aider les radiologues dans leur diagnostic et à la communauté de l'imagerie médicale de pouvoir extraire des informations plus précises. Cette collaboration permet non seulement d'avancer dans les sciences médicales (ex. étude de l'anatomie et physiologie), mais également dans les applications cliniques (ex. détection de maladies et planification de thérapie).
@misc{ling-2012-magazine-article,
title = {L'imagerie médicale: Un domaine d'ingénieurie et de recherche au service de la société},
author = {François Chung},
year = {2012},
date = {2012-01-01},
journal = {L'Ing.},
number = {17},
abstract = {Depuis ses débuts, l'imagerie médicale a pour objectif de fournir aux radiologues des images médicales afin de les aider dans leur diagnostic. Avec l'avancée des techniques d'acquisition, les radiologues se retrouvent à analyser des images de plus en plus complexes et dans des quantités de plus en plus importantes. Du côté de la recherche, cela se traduit par une collaboration entre physique médicale, radiologie et imagerie médicale. Les physiciens ont pour objectif d'améliorer la qualité et la résolution des images médicales. Ces améliorations permettent d'aider les radiologues dans leur diagnostic et à la communauté de l'imagerie médicale de pouvoir extraire des informations plus précises. Cette collaboration permet non seulement d'avancer dans les sciences médicales (ex. étude de l'anatomie et physiologie), mais également dans les applications cliniques (ex. détection de maladies et planification de thérapie).},
howpublished = {L'Ing., 17, pp. 10-12},
keywords = {imagerie médicale, simulation informatique, traitement du patient},
pubstate = {published},
tppubtype = {misc}
}
Depuis ses débuts, l'imagerie médicale a pour objectif de fournir aux radiologues des images médicales afin de les aider dans leur diagnostic. Avec l'avancée des techniques d'acquisition, les radiologues se retrouvent à analyser des images de plus en plus complexes et dans des quantités de plus en plus importantes. Du côté de la recherche, cela se traduit par une collaboration entre physique médicale, radiologie et imagerie médicale. Les physiciens ont pour objectif d'améliorer la qualité et la résolution des images médicales. Ces améliorations permettent d'aider les radiologues dans leur diagnostic et à la communauté de l'imagerie médicale de pouvoir extraire des informations plus précises. Cette collaboration permet non seulement d'avancer dans les sciences médicales (ex. étude de l'anatomie et physiologie), mais également dans les applications cliniques (ex. détection de maladies et planification de thérapie).
@book{lap-2011-book,
title = {Regional appearance modeling for model-based image segmentation: Methodological approaches to improve the accuracy of model-based image segmentation},
author = {François Chung},
isbn = {978-3844322095},
year = {2011},
date = {2011-03-01},
publisher = {Lambert Academic Publishing (LAP)},
address = {Saarbrücken},
abstract = {This thesis presents a novel appearance prior for model-based image segmentation. This appearance prior, denoted as Multimodal Prior Appearance Model (MPAM), is built upon an Expectation–Maximization (EM) clustering of intensity profiles with model order selection to automatically select the number of profile classes. Unlike classical approaches based on Principal Component Analysis (PCA), the clustering is considered as regional because intensity profiles are classified for each mesh and not for each vertex. Comparative results on liver profiles from Computed Tomography (CT) images show that MPAM outperforms PCA-based appearance models. Finally, methods for the analysis of lower limb structures from Magnetic Resonance (MR) images are presented. A first part deals with the creation of subject-specific models for kinematic simulations of the lower limbs. In a second part, the performance of statistical models is compared in the context of lower limb bone segmentation when only a small number of datasets is available for training.},
keywords = {appearance modeling, model-based image segmentation, unsupervised clustering},
pubstate = {published},
tppubtype = {book}
}
This thesis presents a novel appearance prior for model-based image segmentation. This appearance prior, denoted as Multimodal Prior Appearance Model (MPAM), is built upon an Expectation–Maximization (EM) clustering of intensity profiles with model order selection to automatically select the number of profile classes. Unlike classical approaches based on Principal Component Analysis (PCA), the clustering is considered as regional because intensity profiles are classified for each mesh and not for each vertex. Comparative results on liver profiles from Computed Tomography (CT) images show that MPAM outperforms PCA-based appearance models. Finally, methods for the analysis of lower limb structures from Magnetic Resonance (MR) images are presented. A first part deals with the creation of subject-specific models for kinematic simulations of the lower limbs. In a second part, the performance of statistical models is compared in the context of lower limb bone segmentation when only a small number of datasets is available for training.
@article{tvc-2011-academic-journal-article,
title = {Comparison of statistical models performance in case of segmentation using a small amount of training datasets},
author = {François Chung and Jérôme Schmid and Nadia Magnenat-Thalmann and Hervé Delingette},
doi = {10.1007/s00371-010-0536-9},
issn = {0178-2789},
year = {2011},
date = {2011-02-01},
journal = {The Visual Computer (TVC)},
volume = {27},
number = {2},
pages = {141-151},
abstract = {Model-based image segmentation has been extensively used in medical imaging to learn both shape and appearance of anatomical structures from training datasets. The more training datasets are used, the more accurate is the segmented model as we account for more information about its variability. However, training datasets of large size with a proper sampling of the population may not always be available. In this paper, we compare the performance of statistical models in the context of lower limb bones segmentation using Magnetic Resonance (MR) images when only a small number of datasets is available for training. For shape, both priors based on Principal Component Analysis (PCA) and shape memory strategies are tested. For appearance, methods based on intensity profiles are tested, namely mean intensity profiles, multivariate Gaussian distributions of profiles and multimodal profiles from Expectation-Maximization (EM) clustering. Segmentation results show that local and simple methods perform the best when a small number of datasets is available for training. Conversely, statistical methods feature the best segmentation results when the number of training datasets is increased.},
keywords = {model based segmentation, principal component analysis, statistical models},
pubstate = {published},
tppubtype = {article}
}
Model-based image segmentation has been extensively used in medical imaging to learn both shape and appearance of anatomical structures from training datasets. The more training datasets are used, the more accurate is the segmented model as we account for more information about its variability. However, training datasets of large size with a proper sampling of the population may not always be available. In this paper, we compare the performance of statistical models in the context of lower limb bones segmentation using Magnetic Resonance (MR) images when only a small number of datasets is available for training. For shape, both priors based on Principal Component Analysis (PCA) and shape memory strategies are tested. For appearance, methods based on intensity profiles are tested, namely mean intensity profiles, multivariate Gaussian distributions of profiles and multimodal profiles from Expectation-Maximization (EM) clustering. Segmentation results show that local and simple methods perform the best when a small number of datasets is available for training. Conversely, statistical methods feature the best segmentation results when the number of training datasets is increased.
@phdthesis{ph-d-thesis-2011-doctoral-thesis,
title = {Regional appearance modeling for deformable model-based image segmentation},
author = {François Chung},
year = {2011},
date = {2011-01-01},
address = {Sophia-Antipolis},
school = {Doctoral thesis (Ph.D. Thesis), Mines ParisTech, Centre de Mathématiques Appliquées},
abstract = {This thesis presents a novel appearance prior for model-based image segmentation. This appearance prior, denoted as Multimodal Prior Appearance Model (MPAM), is built upon an Expectation-Maximization (EM) clustering of intensity profiles with model order selection to automatically select the number of profile classes. Unlike classical approaches based on Principal Component Analysis (PCA), the clustering is considered as regional because intensity profiles are classified for each mesh and not for each vertex.
First, we explain how to build a MPAM from a training set of meshes and images. The clustering of intensity profiles and the determination of the number of appearance regions by a novel model order selection criterion are explained. A spatial regularization approach to spatially smooth the clustering of profiles is presented and the projection of the appearance information from each dataset on a reference mesh is described.
Second, we present a boosted clustering based on spectral clustering, which optimizes the clustering of profiles for segmentation purposes. The representation of the similarity between data points in the spectral space is explained. Comparative results on liver profiles from Computed Tomography (CT) images show that our approach outperforms PCA-based appearance models.
Finally, we present methods for the analysis of lower limb structures from Magnetic Resonance (MR) images. In a first part, our technique to create subject-specific models for kinematic simulations of lower limbs is described. In a second part, the performance of statistical models is compared in the context of lower limb bones segmentation when only a small number of datasets is available for training.},
keywords = {appearance modeling, model-based image segmentation, unsupervised clustering},
pubstate = {published},
tppubtype = {phdthesis}
}
This thesis presents a novel appearance prior for model-based image segmentation. This appearance prior, denoted as Multimodal Prior Appearance Model (MPAM), is built upon an Expectation-Maximization (EM) clustering of intensity profiles with model order selection to automatically select the number of profile classes. Unlike classical approaches based on Principal Component Analysis (PCA), the clustering is considered as regional because intensity profiles are classified for each mesh and not for each vertex.
First, we explain how to build a MPAM from a training set of meshes and images. The clustering of intensity profiles and the determination of the number of appearance regions by a novel model order selection criterion are explained. A spatial regularization approach to spatially smooth the clustering of profiles is presented and the projection of the appearance information from each dataset on a reference mesh is described.
Second, we present a boosted clustering based on spectral clustering, which optimizes the clustering of profiles for segmentation purposes. The representation of the similarity between data points in the spectral space is explained. Comparative results on liver profiles from Computed Tomography (CT) images show that our approach outperforms PCA-based appearance models.
Finally, we present methods for the analysis of lower limb structures from Magnetic Resonance (MR) images. In a first part, our technique to create subject-specific models for kinematic simulations of lower limbs is described. In a second part, the performance of statistical models is compared in the context of lower limb bones segmentation when only a small number of datasets is available for training.
@conference{miccai-2009-conference-proceeding,
title = {Multimodal prior appearance models based on regional clustering of intensity profiles},
author = {François Chung and Hervé Delingette},
editor = {Guang-Zhong Yang and David Hawkes and Daniel Rueckert and Alison Noble and Chris Taylor},
doi = {10.1007/978-3-642-04271-3_127},
isbn = {978-3-642-04270-6},
year = {2009},
date = {2009-09-01},
booktitle = {MICCAI 2009: Medical Image Computing and Computer-Assisted Intervention},
pages = {1051-1058},
publisher = {Springer Berlin Heidelberg},
address = {London},
series = {Lecture Notes in Computer Science (5762),},
abstract = {Model-based image segmentation requires prior information about the appearance of a structure in the image. Instead of relying on Principal Component Analysis (PCA) such as in Statistical Appearance Models (SAMs), we propose a method based on a regional clustering of intensity profiles that does not rely on an accurate pointwise registration. Our method is built upon the Expectation-Maximization (EM) algorithm with regularized covariance matrices and includes spatial regularization. The number of appearance regions is determined by a novel model order selection criterion. The prior is described on a reference mesh where each vertex has a probability to belong to several intensity profile classes.},
keywords = {covariance matrix regularization, EM classification, intensity profile, order selection},
pubstate = {published},
tppubtype = {conference}
}
Model-based image segmentation requires prior information about the appearance of a structure in the image. Instead of relying on Principal Component Analysis (PCA) such as in Statistical Appearance Models (SAMs), we propose a method based on a regional clustering of intensity profiles that does not rely on an accurate pointwise registration. Our method is built upon the Expectation-Maximization (EM) algorithm with regularized covariance matrices and includes spatial regularization. The number of appearance regions is determined by a novel model order selection criterion. The prior is described on a reference mesh where each vertex has a probability to belong to several intensity profile classes.
@conference{cbm-2009-conference-proceeding,
title = {Subject-specific ligament models: Towards real-time simulation of the knee joint},
author = {Tobias Heimann and François Chung and Hans Lamecker and Hervé Delingette},
editor = {Karol Miller and Poul M F Nielsen},
doi = {10.1007/978-1-4419-5874-7_12},
isbn = {978-1-4419-5873-0},
year = {2009},
date = {2009-09-01},
booktitle = {CBM 2009: Computational Biomechanics for Medicine IV},
publisher = {Springer New York},
address = {London},
abstract = {We present an efficient finite element method to simulate a transversely isotropic non-linear material for ligaments. The approach relies on tetrahedral elements and exploits the geometry to optimize computation of the derivatives of the strain energy. To better support incompressibilty, deviatoric and dilational responses are uncoupled and a penalty term controls volume preservation. We derive stress and elasticity tensors required for implicit solvers and verify our model against the FEBio software using a variety of load scenarios with synthetic shapes. The maximum node positioning error for ligament materials is < 5% for strains under physiological conditions. To generate subject-specific ligament models, we propose a novel technique to estimate fiber orientation from segmented ligament geometry. The approach is based on an automatic centerline extraction and generation of the corresponding diffusion field. We present results for a medial collateral ligament segmented from standard Magnetic Resonance Imaging (MRI) data. Results show the general viability of the method, but also the limitations of current MRI acquisitions. In the future, we hope to employ the presented techniques for real-time simulation of knee surgery.},
keywords = {finite element method, knee, ligament, medical imaging, real-time, simulation},
pubstate = {published},
tppubtype = {conference}
}
We present an efficient finite element method to simulate a transversely isotropic non-linear material for ligaments. The approach relies on tetrahedral elements and exploits the geometry to optimize computation of the derivatives of the strain energy. To better support incompressibilty, deviatoric and dilational responses are uncoupled and a penalty term controls volume preservation. We derive stress and elasticity tensors required for implicit solvers and verify our model against the FEBio software using a variety of load scenarios with synthetic shapes. The maximum node positioning error for ligament materials is < 5% for strains under physiological conditions. To generate subject-specific ligament models, we propose a novel technique to estimate fiber orientation from segmented ligament geometry. The approach is based on an automatic centerline extraction and generation of the corresponding diffusion field. We present results for a medial collateral ligament segmented from standard Magnetic Resonance Imaging (MRI) data. Results show the general viability of the method, but also the limitations of current MRI acquisitions. In the future, we hope to employ the presented techniques for real-time simulation of knee surgery.
ORASIS'09: Congrès des jeunes chercheurs en vision par ordinateur, Association Française pour la Reconnaissance et l'Interprétation des Formes (AFRIF), Trégastel, 2009.
@conference{orasis-2009-conference-proceeding,
title = {Reconstruction 3D des structures anatomiques des membres inférieurs},
author = {François Chung and Jérôme Schmid and Olivier Clatz and Nadia Magnenat-Thalmann and Hervé Delingette},
year = {2009},
date = {2009-06-01},
booktitle = {ORASIS'09: Congrès des jeunes chercheurs en vision par ordinateur},
address = {Trégastel},
organization = {Association Française pour la Reconnaissance et l'Interprétation des Formes (AFRIF),},
abstract = {Dans cet article, nous nous intéressons à la modélisation des structures anatomiques des membres inférieurs telles que les os, les muscles et les tendons. La méthode proposée commence par une acquisition d'images par résonance magnétique (IRM) durant laquelle les membres inférieurs d'un sujet sont scannés. Des modèles 3D sont ensuite générés après une segmentation manuelle des structures anatomiques. Cependant, la surface des modèles générés n'est pas lisse. De plus, les modèles ne sont pas attachés alors qu'ils devraient l'être anatomiquement. Nous décrivons donc les différentes étapes pour contraindre les modèles à être corrects au niveau anatomique et nous discutons de leur validation. L'objectif de cette méthode est de pouvoir réutiliser ces modèles dans des méthodes de segmentation automatique.},
keywords = {IRM, membres inférieurs, modélisation 3D, segmentation, traitement d’images},
pubstate = {published},
tppubtype = {conference}
}
Dans cet article, nous nous intéressons à la modélisation des structures anatomiques des membres inférieurs telles que les os, les muscles et les tendons. La méthode proposée commence par une acquisition d'images par résonance magnétique (IRM) durant laquelle les membres inférieurs d'un sujet sont scannés. Des modèles 3D sont ensuite générés après une segmentation manuelle des structures anatomiques. Cependant, la surface des modèles générés n'est pas lisse. De plus, les modèles ne sont pas attachés alors qu'ils devraient l'être anatomiquement. Nous décrivons donc les différentes étapes pour contraindre les modèles à être corrects au niveau anatomique et nous discutons de leur validation. L'objectif de cette méthode est de pouvoir réutiliser ces modèles dans des méthodes de segmentation automatique.
Nadia Magnenat-Thalmann; Jian J Zhang; David D Feng (Ed.): Recent Advances in the 3D Physiological Human (3DPH), pp. 3-19, Springer-Verlag, London, 2009, ISBN: 978-1-84882-565-9.
@inbook{3dph-2009-book-chapter,
title = {Musculoskeletal simulation model generation from MRI datasets and motion capture data},
author = {Jérôme Schmid and Anders Sandholm and François Chung and Daniel Thalmann and Hervé Delingette and Nadia Magnenat-Thalmann},
editor = {Nadia Magnenat-Thalmann and Jian J Zhang and David D Feng},
doi = {10.1007/978-1-84882-565-9},
isbn = {978-1-84882-565-9},
year = {2009},
date = {2009-02-01},
booktitle = {Recent Advances in the 3D Physiological Human (3DPH)},
pages = {3-19},
publisher = {Springer-Verlag},
address = {London},
abstract = {Today, computer models and computer simulations of the musculoskeletal system are widely used to study the mechanisms behind human gait and its disorders. The common way of creating musculoskeletal models is to use a generic musculoskeletal model based on data derived from anatomical and biomechanical studies of cadaverous specimens. To adapt this generic model to a specific subject, the usual approach is to scale it. This scaling has been reported to introduce several errors because it does not always account for subject-specific anatomical differences. As a result, a novel semi-automatic workflow is proposed that creates subject-specific musculoskeletal models from Magnetic Resonance Imaging (MRI) datasets and motion capture data. Based on subject-specific medical data and a model-based automatic segmentation approach, an accurate modeling of the anatomy can be produced while avoiding the scaling operation. This anatomical model coupled with motion capture data, joint kinematics information and muscle-tendons actuators is finally used to create a subject-specific musculoskeletal model.},
keywords = {computer simulation, lower limb, medical image processing, motion capture},
pubstate = {published},
tppubtype = {inbook}
}
Today, computer models and computer simulations of the musculoskeletal system are widely used to study the mechanisms behind human gait and its disorders. The common way of creating musculoskeletal models is to use a generic musculoskeletal model based on data derived from anatomical and biomechanical studies of cadaverous specimens. To adapt this generic model to a specific subject, the usual approach is to scale it. This scaling has been reported to introduce several errors because it does not always account for subject-specific anatomical differences. As a result, a novel semi-automatic workflow is proposed that creates subject-specific musculoskeletal models from Magnetic Resonance Imaging (MRI) datasets and motion capture data. Based on subject-specific medical data and a model-based automatic segmentation approach, an accurate modeling of the anatomy can be produced while avoiding the scaling operation. This anatomical model coupled with motion capture data, joint kinematics information and muscle-tendons actuators is finally used to create a subject-specific musculoskeletal model.
Emili Hernàndez; Pere Ridao; Marc Carreras; David Ribas; Narcís Palomeras; Andres El-Fakdi; François Chung; Toni Almohaya; Xavier Ribas; Guillermo García de Marina; Joan Massich; Natàlia Hurtós
@conference{ccia-2006-conference-proceeding,
title = {ICTINEU AUV, un robot per a competir},
author = {Emili Hernàndez and Pere Ridao and Marc Carreras and David Ribas and Narcís Palomeras and Andres El-Fakdi and François Chung and Toni Almohaya and Xavier Ribas and Guillermo García de Marina and Joan Massich and Natàlia Hurtós},
year = {2006},
date = {2006-10-01},
booktitle = {CCIA 2006: Congrés Català d'Intel.ligència Artificial},
address = {Perpignan},
organization = {Associació Catalana d’Intel.ligència Artificial (ACIA),},
abstract = {SAUC-E (Student Autonomous Underwater Challenge - Europe) és la primera competició en robòtica submarina per a estudiants que es dur a terme a Europa. Inspirada en les competicions americanes, per a competir en el SAUC-E, un grup d'estudiants han de dissenyar i construir un robot capaç de dur a terme una missió predefinida per l'organització, de forma totalment autònoma. Un grup d'estudiants i professors de la Universitat de Girona (UdG) hem decidit afrontar aquest repte construint el robot ICTINEU AUV. Aquest robot aprofitarà eines de maquinari i programari desenvolupades els darrers anys en el nostre laboratori. Aquest article presenta el robot que estem desenvolupant i presenta resultats preliminars (simulació en temps real) de la seva arquitectura de control, mostrant la viabilitat de la proposta.},
keywords = {competició d’estudiants, disseny de robot, temps real, vehicle submarí autònom},
pubstate = {published},
tppubtype = {conference}
}
SAUC-E (Student Autonomous Underwater Challenge - Europe) és la primera competició en robòtica submarina per a estudiants que es dur a terme a Europa. Inspirada en les competicions americanes, per a competir en el SAUC-E, un grup d'estudiants han de dissenyar i construir un robot capaç de dur a terme una missió predefinida per l'organització, de forma totalment autònoma. Un grup d'estudiants i professors de la Universitat de Girona (UdG) hem decidit afrontar aquest repte construint el robot ICTINEU AUV. Aquest robot aprofitarà eines de maquinari i programari desenvolupades els darrers anys en el nostre laboratori. Aquest article presenta el robot que estem desenvolupant i presenta resultats preliminars (simulació en temps real) de la seva arquitectura de control, mostrant la viabilitat de la proposta.
David Ribas; Narcís Palomeras; Xavier Ribas; Guillermo García de Marina; Emili Hernàndez; François Chung; Natàlia Hurtós; Joan Massich; Toni Almohaya; Josep Vila
@techreport{udg-2006-technical-report,
title = {ICTINEU AUV takes the challenge},
author = {David Ribas and Narcís Palomeras and Xavier Ribas and Guillermo García de Marina and Emili Hernàndez and François Chung and Natàlia Hurtós and Joan Massich and Toni Almohaya and Josep Vila},
year = {2006},
date = {2006-01-01},
address = {Girona},
institution = {Universitat de Girona (UdG),},
abstract = {Within 1857-1868, Narcis Monturiol developed the first Spanish submarine. Almost one hundred and fifty years later, a pioneer team of students of the University of Girona (UdG) decided to design and develop an Autonomous Underwater Vehicle (AUV) to face the Student Autonomous Underwater Challenge – Europe (SAUC-E). Our robot, called ICTINEU AUV, is the result of a multidisciplinary project involving undergraduate, graduate and Ph.D. students of industrial engineering and computer science. The prototype has evolved from the initial Computer Aided Design (CAD) model to become an operative Autonomous Underwater Vehicle (AUV) in the short period of seven months. The open frame and modular design principles, together with the compatibility with other robots previously developed at the lab, have been the main design philosophy. Hence, at the robot core, two networked computers give access to a wide set of sensors and actuators. The Gentoo/Linux distribution, with a 2.6 pre-emptive kernel, has been chosen as the onboard operating system. The real-time POSIX, together with the ACE/TAO CORBA-RT ORB, have been extensively used to develop the control architecture as a set of distributed objects with soft real time capabilities. Common software engineering practises have been applied to ensure software reliability including Unified Modelling Language (UML) design, extensive documentation using DoxyGen and the use of a Code Version Server (CVS) to handle the sharing of multiple code versions. Finally, in order to reduce the development time, concurrent engineering techniques based on Hardware In the Loop (HIL) simulation have been applied to overlap the hardware and software design and development.},
keywords = {autonomous underwater vehicle, real-time, robot design, student competition},
pubstate = {published},
tppubtype = {techreport}
}
Within 1857-1868, Narcis Monturiol developed the first Spanish submarine. Almost one hundred and fifty years later, a pioneer team of students of the University of Girona (UdG) decided to design and develop an Autonomous Underwater Vehicle (AUV) to face the Student Autonomous Underwater Challenge – Europe (SAUC-E). Our robot, called ICTINEU AUV, is the result of a multidisciplinary project involving undergraduate, graduate and Ph.D. students of industrial engineering and computer science. The prototype has evolved from the initial Computer Aided Design (CAD) model to become an operative Autonomous Underwater Vehicle (AUV) in the short period of seven months. The open frame and modular design principles, together with the compatibility with other robots previously developed at the lab, have been the main design philosophy. Hence, at the robot core, two networked computers give access to a wide set of sensors and actuators. The Gentoo/Linux distribution, with a 2.6 pre-emptive kernel, has been chosen as the onboard operating system. The real-time POSIX, together with the ACE/TAO CORBA-RT ORB, have been extensively used to develop the control architecture as a set of distributed objects with soft real time capabilities. Common software engineering practises have been applied to ensure software reliability including Unified Modelling Language (UML) design, extensive documentation using DoxyGen and the use of a Code Version Server (CVS) to handle the sharing of multiple code versions. Finally, in order to reduce the development time, concurrent engineering techniques based on Hardware In the Loop (HIL) simulation have been applied to overlap the hardware and software design and development.
@mastersthesis{m-sc-thesis-2005-masters-thesis,
title = {Supervision centralisée d'infrastructures distantes en réseaux avec gestion des alarmes et notification des alertes},
author = {François Chung and Jonathan Lousse},
year = {2005},
date = {2005-06-30},
address = {Brussels},
school = {Master’s thesis (M.Sc. Thesis), Institut Supérieur Industriel de Bruxelles (ISIB)},
abstract = {Pour assurer la disponibilité permanente de leur infrastructure informatique, les entreprises ont rapidement compris que la supervision était devenue une ressource-clé. Bon nombre de logiciels de supervision existent mais, parmi tout ceux-là, Nagios est très certainement le logiciel Open Source le plus répandu et également le plus suivi par la communauté de développeurs.
La supervision a trois objectifs principaux: prévenir les incidents sur le réseau par extrapolation des données fournies, agir rapidement dès qu'un système est noté en erreur et permettre l'analyse "post mortem" d'un problème grâce aux informations collectées.
Bien que Nagios soit reconnu comme difficile à installer et à configurer, il possède un nombre impressionnant de fonctionnalités telles que la surveillance de services réseau (SMTP, POP3, HTTP, NNTP, PING, ...), la surveillance de ressources machine (charge du processeur, utilisation du disque, ...) et la conception simple des "plugins" qui permet à des utilisateurs de développer facilement leur propre service de vérification.
Dans le cadre de ce travail de fin d'études (TFE), nous avons mis en place une supervision centralisée de l'infrastructure informatique de EASI, une gestion d'alarmes en cas de problèmes sur le réseau et une notification des alertes par email. Pour effectuer cela, nous avons non seulement utilisé les plugins standards de Nagios mais nous avons également développé nos propres plugins.
De plus, nous avons créé un mode multi-utilisateurs avec une authentification via l'annuaire LDAP (Lightweight Directory Access Protocol) de EASI et nous avons sécurisé les échanges d'informations entre les serveurs Nagios par un système de clés publiques et privées. L'objectif à terme est d'étendre cette architecture à l'ensemble des clients de EASI.},
keywords = {serveur central, surveillance réseau et logicielle, système de supervision},
pubstate = {published},
tppubtype = {mastersthesis}
}
Pour assurer la disponibilité permanente de leur infrastructure informatique, les entreprises ont rapidement compris que la supervision était devenue une ressource-clé. Bon nombre de logiciels de supervision existent mais, parmi tout ceux-là, Nagios est très certainement le logiciel Open Source le plus répandu et également le plus suivi par la communauté de développeurs.
La supervision a trois objectifs principaux: prévenir les incidents sur le réseau par extrapolation des données fournies, agir rapidement dès qu'un système est noté en erreur et permettre l'analyse "post mortem" d'un problème grâce aux informations collectées.
Bien que Nagios soit reconnu comme difficile à installer et à configurer, il possède un nombre impressionnant de fonctionnalités telles que la surveillance de services réseau (SMTP, POP3, HTTP, NNTP, PING, ...), la surveillance de ressources machine (charge du processeur, utilisation du disque, ...) et la conception simple des "plugins" qui permet à des utilisateurs de développer facilement leur propre service de vérification.
Dans le cadre de ce travail de fin d'études (TFE), nous avons mis en place une supervision centralisée de l'infrastructure informatique de EASI, une gestion d'alarmes en cas de problèmes sur le réseau et une notification des alertes par email. Pour effectuer cela, nous avons non seulement utilisé les plugins standards de Nagios mais nous avons également développé nos propres plugins.
De plus, nous avons créé un mode multi-utilisateurs avec une authentification via l'annuaire LDAP (Lightweight Directory Access Protocol) de EASI et nous avons sécurisé les échanges d'informations entre les serveurs Nagios par un système de clés publiques et privées. L'objectif à terme est d'étendre cette architecture à l'ensemble des clients de EASI.
