BIO Medical research Imaging SemanTic data management – Biomist

The project is managed with the Scrum methodology adapted to the multidisciplinary and geographically disparate structure of the consortium. A requirement backlog is managed with the Agilefant tool, 4 to 5 sprints are scheduled each year.
In order to define precisely the needs in the on the neuro-functional imaging field, a set of workshops with researchers from the GIN labs are organized, allowing to describe the “users stories” needed for the definition of the sprints. In the same way, the definition of the technical infrastructure is started in order to deploy the necessary IT resources.
The refinement of user stories, of the existing data models and the analysis of existing ontologies in the neurosciences domain enabled us to define a consistent data model deployed on the PLM software, to define classifications compatible with the vocabulary described by the domain ontologies and to deploy workflows and associated image processing programs on computing grids.
The most general user story asks for a easily apprehensible user interface enabling the user to define a request by the way of existing classifications, to present the result (e.g. a set of subjects from existing studies), to start a processing on a computing grid and to analyze the resulting brain connectivity graphs. All these steps require the user to manipulate a complex set of data representation (classifications, data model, brain connectivity graphs, large network comparison,…), this brought us to develop a new definition for multi-dimensional and multivariate configurable graphs that allows to describe the evolution in several dimensions (time, subject, processing, versions,…) of a graph that represent the interaction between brain regions.

One year from the beginning of the project we have the following results:
- The internal communication tools (backlog management, document sharing, teleconference, source code management, continuous integration platform…) and external communication (web site) have been deployed.
- The consortium agreement is being signed.
- A first set of requirements on the general process have been collected, refined and analyzed concerning image acquisition, study & subject definition, processing, scientific papers…
- The technical architecture of the project: computers, licenses, OS,.. have been acquired and deployed.
- The data model has been designed by comparison with existing models: GIN data model, XNAT, OntoNeuroLOG …
- Technical workflows has been defined: acquisition, storage, links with computing grids.
- Processing workflows are being tested.
- Data migration and alignment to the new model from the existing GIN database is done.
- Relations have been established with the projects Codde (ANR-13-CORD-0017), NeuroLOG (ANR-06-TLOG-024), i-Share (http://www.i-share.fr).
- Several software development frameworks for graph representation have been tried (Gephi, Cytoscape, Tulip), however the graph data model used is not comprehensive enough for our needs. We have then developed our own graph representation that may be adapted in order to be compatible with the Gexf format (understood by existing frameworks), this representation allows us to define multi-dimensional and multivariate configurable graphs. The tools for manipulation and presentation of this representation are being developed.
- User interface for acquisition and request definition integrated with the PLM software are being developed.
- Two scientific papers have been accepted.

This first year has been dedicated to the deployment of the fundamental needs for the project : deployment of the technical infrastructure, integration of the image processing workflows, definition and deployment of the PLM tool, its data model and classifications, requirement acquisition and analysis, and definition of the basic user interfaces. Currently, the main obstacle to general usage is that the user interface is too “technical”.
The definition of a user interface unifying in a same paradigm the presentation and manipulation of the complex set of data representation that is used throughout the main process of a neurofunctional study is the main objective of the second year of the project.
The objective for the end of the project is to validate all the concepts and tools developed by taking part in an actual study where the acquired data will used in parallel between the Biomist architecture and process and the existing process and comparing the usability and performance of both systems.

ALLANIC, Marianne, DURUPT, Alexandre, JOLIOT, Marc, EYNARD, Benoît, BOUTINAUD, Philippe. Towards a data model for PLM application in Bio-Medical Imaging. In: Proceedings of TMCE, 2014.
ALLANIC, Marianne, BRIAL, Thierry, DURUPT, Alexandre, JOLIOT, Marc, BOUTINAUD, Philippe, EYNARD, Benoît. Towards an enhancement of relationships browsing in mature PLM systems. In : Product Lifecycle Management for Society. Springer Berlin Heidelberg, 2014.

The BIOMIST (BIO Medical research Imaging SemanTic data management) project is proposed for thematic axis n°2 of the Contint 2013 Call for Proposal: from content to knowledge and big data. The project objective is to provide researchers in the field of biomedical imaging (BMI) with an efficient information system in order to help them make optimal use of their data during their research activities on large group of subjects, and to enable the reuse of available data for clinic and fundamental research in a context and for a purpose other than the one for which they have been acquired.

The partners for this project are Cadesis (a R&D SME specialized in the integration of Information Systems for the industry), Groupe d'Imagerie Neurofonctionnelle (GIN - UMR 5296) a core member of the LabEx TRAIL "Investissements d'avenir", Laboratoire Roberval (UMR 7337 - Université Technologique de Compiègne) and Institut Charles Delaunay (ICD - UMR 6279 - Université Technologique de Troyes).

With the GIN research team we will focus on the neuro-functional imaging field as a testbed for our project. Besides the images (2D, 3D, 4D), we will manage every type of data required such as demographic data, behavioral test results and as a new topic in this field: genetic information. A special focus will be on both intra and inter subject analysis definition and results with respect to the material used for scientific publications. Our project will aim to manage not only documents but, more significantly, the specific concepts used in neuro-functional analysis such as cognitive stimulation paradigms, processing tasks, behavioral test definitions… and all the semantic relationships that may exist between them.

We aim to provide methodologies and tools to manage the growing amount, complexity and provenance of BMI data and to consider, not only the data, but also their usage, their different representation and their interpretation in the context of neuro-functional research. We propose to use a proved foundation used in traditional engineering to answer the basic requirements for BMI data management: Product Lifecycle Management (PLM) solutions. Since our application domain is a research environment, it demands more flexibility than what is currently available from PLM solutions; we then propose to use Knowledge Management (KM) techniques to enhance reuse and traceability in the usage of BMI data in a complex and changeable environment. Furthermore we will provide and integrate visualization and analysis tools that enable users to make hypothesis, intuitively discover and compare patterns, and isolate structure singularities in graph representation of data that could be used for semantic relationships or brain connectivity graphs (a specific neuro-functional representation).

The results of the project will be a working prototype deployed, validated and used by the GIN team. The technology transfers between academic partners and Cadesis during the project are expected to lead to exploitable results such as the availability of a BioMedical Imaging module for PLM solutions, the enhancement of Knowledge-based reuse methodologies and the development of graph visualization tools and algorithms that could be applied to a wide range of domains.