The field of neuroimaging research requires the ability to share, reuse and compare image processing tools coming from different laboratories. However, sharing treatment as services and composing them as workflows, is usually difficult and a complex task. This is due in most cases to the heterogeneity of services and platforms with regards to their conception and their implementation. We work within the NeuroLOG project, which aims at developing a middleware to federate data repositories and to facilitate the sharing and reuse of processing tools to analyze the shared images. It adopts an ontological approach for data and tools mediation and for sharing resources. This work aims to provide tools mediation to enhance the sharing and composition of image processing tools and provide non-specialist and expert users of neuroimaging field with an ergonomic and easy to use composition platform. We have chosen to use the Semantic Web techniques to address the various problems of resource interoperability and consistency. The first proposed solution is based on an extension of the OWL-S framework. It has been adapted to the various web services of our neuroimaging platform. We finally concluded that services that haven't the WSDL standard as descriptor could not be chained as workflow. So, we have proposed a new approach to compose image processing tools. It is based on a new ontological model for service composition that meets the requirements of the neuroimaging domain and the constraints of our domain ontology OntoNeuroLOG and addresses the various problems encountered in the development of the first approach. This work led to solve the two major problems in the composition of services; the heterogeneity of services descriptors and the interoperability of services according to the constraints within the NeuroLOG platform.
