Proceedings of the 3 rd International Conference– EMMIT 2007 1 A New Paradigm in Biomedical Data Discovery and Multimodal Workflows Adina Riposan 1 , Ian Taylor 2 , David R. Owens 3 , and Edward C. Conley 4 1 Military Technical Academy, and Contact Net Ltd, Bucharest, Romania, adina.riposan@contactnet.ro 2 School of Computer Science, Cardiff University, UK, and Center for Computation and Technology, LSU, USA, Ian.J.Taylor@cs.cardiff.ac.uk 3,4 Diabetes Research Unit (DRU), School of Medicine, Cardiff University, UK and Diabetic Retinopathy Screening Service for Wales (DRSSW), owensdr@cardiff.ac.uk , e.c.conley@cs.cardiff.ac.uk In this paper we apply an alternative search mechanism to biomedical audio/visual and spectral data discovery using distributed peer-to-peer techniques. We describe the underlying architecture, distributed database, along with the complex search algorithms based on multimodal workflows, aggregating index and content information. We exemplify the utilization of such a distributed infrastructure in diabetic retinopathy research and clinical trials, for the early detection and prevention of retinal disease and investigational drug discovery. The peer-to-peer toolkit itself provides a platform that enables users to build search algorithms by combining components into workflows that are executed across the peers on the network using industry standards such as Web services and SOAP. An aspect of this framework is that it contains a mobile implementation capable of allowing the control and monitoring of searches from mobile devices, as well as remote workflow management by the remote control of the enactment engine from a mobile user interface. Keywords Biomedical imaging, Diabetic Retinopathy, Grid technology, Peer-to-Peer, Search engine 1. Introduction Cardiff University has been working on a generic infrastructure, codenamed the Alchemist, to create a framework that provide a P2P layer for supporting pluggable network discovery and caching overlays coupled with the ability to execute distributed workflows [1]. The framework is being built on an existing middleware system, called WSPeer [2], which provides a SOAP messaging layer (using Web or WS-RF Services) within a P2P network that supports a super-peer topology of rendezvous or advert caching peers to support the scalability of the discovery and access to information on the network as a whole. This system already interfaces with existing Grid middleware (e.g. Globus) through Web Services interfaces but is also able to provide access to such capabilities within a decentralised environment. For example, rather than relying on centralised discovery mechanisms e.g. UDDI or similar, we can search super peers for WSDL files to provide access to the distributed services. In this way, the network can cope with far more transient participants, it supports changing roles