Section: New Results
Management of Information and Semantic Processing
Participants : Bacem Wali, Daniel Garcia-Lorenzo, Franck Michel, Franck Michel, Farooq Ahmad, Bernard Gibaud, Christian Barillot.
Introduction
A better sharing of resources (by "resources" we denote both data and image processing tools) is one of the keys for the success of future research in the medical imaging domain. Especially, the discovery and the validation of new imaging biomarkers for the diagnosis of neurological diseases and the monitoring of their evolution and treatment highly depends on collaborative work associating expert centers with similar or complementary skills. Suitable information infrastructures must be provided to support such collaborative work, enabling an easy sharing of images and associated data and a flexible sharing and re-use of processing tools available at the different sites. Tradionnally, sharing can be envisaged according to two different models, a "centralized" one and a "federated" one. Our assumption is that the latter is much more suited to the biomedical research field, especially due to the legitimate wish of involved organisations to define and control how their data should be organized and shared. One of the most critical aspects of the design of such federated systems is the definition of common semantics of shared information. Ontologies and semantic web technologies can provide powerful solutions to this problem. Ontologies provide both a reference vocabulary and explicit definitions (using formal semantics) of entities, which can be used to reason about real world entities, classify them, and retrieve them from data repositories. One of our basic assumptions is that such technologies should be used in the neuroimaging domain to make explicit the nature and content of the images and relate them to other kinds of data : biological data, clinical data, neuropsychological data, genomic data, etc.
Neurolog project: Sharing of data and sharing of processing tools in neuroimaging
Our participation in Neurolog concerned primarily three workpackages : WP1 "data distribution", WP2 "ontologies and semantic processing" and WP5 "applicative testbests". Regarding the first, our major achievement was the implementation of a middleware (called Neurolog server) to share the data as well as the processing tools. The specific contribution of Visages concerned more specifically the design of the "data manager" and "metadata manager". Concerning the second, two major results were obtained. A first version of the OntoNeurolog ontology was delivered, covering entities such as: Study, Center, Examination, Subject, Dataset Acquisition, Datasets, Dataset expression, Files, Data Processing etc. This first version was made available for use by the other work packages, especially WP1 (development of the Metadata Manager), WP5 (development of the Neurolog Client) and for the development of the InriaNeuroTk software (a toolkit to populate local image databases). This first version of the ontology was also communicated to our colleagues of the BIRN project in the USA (Dr Jessica Turner, University of California, Irvine). A second version was issued in september 2009, including an ontology of "Instruments" (neuropsychological tests, behavioural tests, neuroclinical scales, as well as the various scores resulting of their use). This is very important since most neuroimaging studies aim at correlating observations made on images (based on image processing) with (quite) objective clinical observations made by healthcare professionals using such instruments. Important work was also done about modeling MR protocols and sequences, but still not delivered. The second significant result is the implementation of a semantic query software, based on the METAMORPHOSES and CORESE software packages. This software is currently being deployed according to a client-server architecture (semantic module implemented in the Neurolog server and accessed through web services). Finally, we participated in the functional specification and to the development of the GUI of the NeuroLOG client application. In particular, a "browse metadata" GUI tab allows users to browse through metadata of the federated database along a predefined (though customizable) browsing tree, using different search criteria. Ultimately, the browsing allows to collect datasets into the user cart, that may be used to: (1) explore details of each dataset object, (2) download data files from a specific dataset expression (i.e. format, such as DICOM or NifTi) and pass them to the image viewer (developed by Visioscopy), (3) use dataset expressions as inputs for processing tools invocation.
Neurolog is a collaborative project, supported by ANR (Agence National de la Recherche), through grant ANR-06-TLOG-024. The partners with whom we have the tightest relations are: I3S (Sophia) and Business Objects, for WP1; MIS (Amiens), GIN (Grenoble) and PitiŽ-Salptrire (Paris) for WP2; GIN (Grenoble) and PitiŽ-Salptrire (Paris) for WP5.
