Ignacio Blanquer explained how the Grid in general can provide access to distributed information seamlessly and how it is able to access advanced applications for data processing as well as high-performance computing resources. As such it functions as a framework for collaborative work in a secure, efficient, transparent and robust way, and not interfering particular administrative policies. Yet, the Grid is not merely an interface to high-performance computers, an application service provider nor an unlimited source of resources. According to the speaker, the Grid should not be considered to be an ideal platform for standard parallel computing applications only, nor as a job queue management system either.
Instead, a Grid can pay excellent services whenever a coherent and secure access to distributed data is needed or when data processing needs to be done in each end. If a programme has to be run many times for a panoply of parameters or for use in large coarse-grain parallel applications, the Grid is also of great value.
Now eHealth, as the speaker stated, is specifically dealing with using information and communication technologies (ICT) to develop an intelligent environment that enables ubiquitous management of the citizens' health status, assisting health professionals in coping with major challenges, and integrating the advances in health knowledge into clinical practice. How can the Grid with its specific features contribute to solve this type of issues?
First, we have to look at what physicians need exactly, according to Ignacio Blanquer. From the individualised point of view, they need access to all the relevant health information of a patient regardless to the place where it is stored. They also want to use computer-aided tools to interpret the patient-specific data and to define the most suitable therapy from the diagnostic. From an epidemiologic point of view access to all the relevant health knowledge of populations is required as well as support in the research for new therapies and drugs.
Ignacio Blanquer stressed the importance of an ubiquitous, quick and secure transference of medical information. This type of information is large and still increasing in size, consisting of multimedial data which is relevant on the long term and critically confidential. The legal regulations at a national level hamper the integration of large repositories of personal data. On top of this, it is nearly impossible to store all the data outside the hospital borders with plain access to the whole set of personal data. It is neither easy to provide access from the Internet, just like that, or to replicate data, even temporally.
In the medical sector, it is vital to have consolidated access to the data, according to the speaker. As such, the different isles of information including HIS,RIS, LIS, PACS and primary information systems have to be interconnected. At the level of the different data formats being used, incompatibilities have to be solved. Medical data indeed is heterogeneous in format and multimedial in nature which increases complexity. In addition, the information has to be complete. This is not evident since some data potentially is incomparable.
To store and process medical images, vital signs monitoring and genetic information, large computing power is needed. The information processing requires large computing resources for keeping statistics of patient records, for knowledge extraction using data mining and for simulation by means of complex biomedical models. Therefore, parallel computing should be considered, as the speaker explained to the audience.
In addition, since health care is performed around the clock, seven days a week, systems with pervasive access and an absolute fault tolerance are indispensable. However, it is difficult to guarantee an "always on" application with a 100 percent quality of service, as the speaker stated. He also insisted that during the past years, a lot of new applications have emerged in medicine such as computer-aided diagnosis; accurate therapy planning; large-scale knowledge discovery; patient-customized analysis; large scale post-genomics and proteomics research; and biological simulation.
Hence the strengths and weaknesses of Grids used for health care have to be analysed. Ignacio Blanquer came up with a host of driving factors at the structural, economical and technological level. Grids allow to generate networks of centres which co-operate at achieving a common objective. Staff is able to share experiences and even personnel. The distributed architecture of the Grid responds well to the idea of health care provision. The concept of the virtual organisation is crucial in Grid philosophy and perfectly matches the structure of health care users. As for the infrastructure, currently high-bandwidth networks are available to both public and private health care centres. There is also an open-mindedness towards research and advanced IT combined with a strong concern to satisfy the patient.
When talking economics, we have to realize that a service-oriented business, as offered by the Grid, is optimal for health care. The speaker also mentioned a better use of resources, the free maintenace of tasks, an improved global IT organisation, scalable costs, and a large and consolidated IT business within the health care organisation. From a technical point of view, the Grid is able to provide security, interoperability and resource sharing. Filters can be executed locally to transform data into a homogeneous format. The Grid preserves the local administration and has low requirements on the user side while being robust at the same time. The Grid provides access to distributed and replicated databases, multiple computing resources and can reallocate tasks.
Grids are not all good news though. The speaker also highlighted some important barriers which have to be taken into account. Physicians have a high legal and moral responsibility towards their patients when medical risks are involved using Grid-related therapy. There is also a shortage of time for physicians to get properly acquainted with the development and testing of Grid-enabled applications and difficulties to overcome when learning new procedures. In addition, medical data are subject to legal requirements with respect to confidentiality. A Grid-enabled service model involves management in ownership of critical client functions, high service-level requirements, as well as a strict policy regarding long-time storage and privacy regulations of health records.
In search for an answer to the question what is available at present in Grid technology and which Grid applications actually do exist in health, Ignacio Blanquer referred to the Globus Toolkit which is the most widely used middleware to date. Globus provides protocols to support resource monitoring and discovery service; job submission and monitoring; and data transfer. It also has the infrastructure to federate distributed resources as a support for access control and to care of wide area security mechanisms.
The Globus Grid is service-oriented, a service consisting of an interoperability protocol and an application behaviour. Its Web services are applications identified by a URL and support interactions by using XML and Web protocols. It is defined in WSDL, discovered with UDDI and interoperable due to SOAP. This Grid together with its Web services is known as the Open Grid Service Architecture or OGSA Grid Service, as the speaker stated. It provides the standard interface for describing Grid services: OGSA defines the semantics of a Grid service but not the service application behaviour. The Open Grid Service Infrastructure (OGSI) is a technical specification of OGSA, which facilitates co-ordinated resource sharing beyond the scientific community boundaries and GT3 is an implementation of OGSI.
The speaker predicted a proliferation of Web and Grid services. The compatibility in Grid applications is focused on their syntactic compatibility. In many cases, the data is readable but not understandable by computers, so an effort is being put on the semantic interpretation and compatibility by giving a definition of the semantics of the services as well as of the ontologies, which refer to capabilities rather than interfaces. So service interoperability through a standard syntactic interface (OGSI) and resource interoperability by a semantic metadata description (Semantic Web) enable an interoperable Grid service with semantic metadata description.
As far as a Grid for health is concerned, one must understand that its general business model is more complex than that of a health Application Service Provider. High throughput or HPC Grids are clear, as Ignacio Blanquer explained, but health Grids are more closely related to data. Although data storage is a responsibility for hospitals, many business opportunities can arise from data sharing and processing applications. For the moment, health Grids are still in a very early stage. Other Grids, for instance for high energy physics, financial, or aerospace applications, are more advanced.
In order to make progress in health, applications and projects should be brought closer to the physicians. The speaker here was thinking of medical imaging applications such as mammography or content-based retrieval; therapy like for example radiotherapy simulation; and surgery planning for neurological, vascular, maxillo-facial, cardiac, or aesthetical procedures. Other health care applications and projects which ought to be considered, can be found in the domains of simulation and biocomputing. The speaker mentioned computational fluid dynamics for drug delivery and blood flow; electrical and mechanical applications; imaging modalities for PET and SPECT; biomolecular modelling; and genomics.
An exemplary project in the biomedical area is EGEE - Enabling Grids for E-science in Europe. Its major aim is to deploy a production Grid infrastructure at the European level. Applications to be run on the EGEE Grid are oriented to high energy physics; biomedicine and medical IT; and other application areas. According to the speaker, EGEE will foster the deployment of health Grid applications by providing a powerful and robust Grid platform for the different applications; suppport for migrating the applications to the Grid; European visibility; and access to a large community of medical users.
In conclusion, Ignacio Blanquer gave an overview of the conditions for health Grids to prosper in the future. The key for success in health Grid application deployment is medical relevance. The way to succeed can be blocked if a health application includes an overhead on medical practice; is not useful in general practice; is not sufficiently reliable, robust nor secure; or is changing the medical procedures fundamentally. A health Grid application needs three collaborating parties: physicians, engineers, and technology providers.
Hence, a killer application in Grid for health has to respect data privacy policies; solve a problem of resource sharing; require computing; be user-driven, OGSA-compliant, and robust; and gradually substitute the current practice. The speaker in depth explained each of these conditions to the audience. With respect for data privacy policies he meant that Grid nodes should be installed within medical data networks. The data should be stored encrypted. Information that flows across the Grid should be filtered to show only the relevant and authorized data. In addition, one should consider the legal issues of European and national directives whereas confidentiality, integrity, authorization and confidentiality should be provided by the Grid.
When it comes to solving a problem of resource sharing, access to distributed medical data should be integral and as complete as possible. The data must be federated and access policies must ensure a uniform and user-friendly access to the data. In addition, a global ontology should be defined. The health Grid application in contrast with distributed databases requires computing. The work has to imply parallelism but no large data distribution. The tasks have to be performed on local databases to raise a global result; the Grid has to distribute different experiments, called the parametrical approach; or distribute the data among resources; or any combination of these three.
With OGSA-compliant, the speaker referred to a killer application as being based on Grid services while integration in standard ICT health applications must be considered. The Grid system is robust when it is fault tolerant and pervasive and when it shows a high degree of repeatability in the results. A killer application is user-driven not by solving an infrequent or particularly complex problem but by solving a daily practitioners' problem. It must also involve users from the very beginning. The health Grid application will gradually substitute the current practice if the implementation is progressive; if the contingency plans are defined; and if the interface, procedures and forms are as similar as possible to standard and confident medical practice.
Ignacio Blanquer finished his talk by stating that he considered ICT for health to be a good candidate for the use of Grid because it is mature and because the current status of Grid technologies is about to reach the desired level of maturity. The problems of health ICT are different from other typically "computing intensive" Grids, however. In any case, successful applications need the commitment and guidance of medical users. The infrastructure of Grids has to pay special attention to hospitals where the data is located. A high degree of co-ordination is needed, as well as a considerable number of success stories. In this regard, medical imaging and biocomputation form application areas in health with a large chance of success at the short or medium term.