The @neurIST project: towards understanding cerebral aneurysms

As for almost every form of knowledge, the volume of data describing human disease processes, including our understanding, diagnosis, and management of them, is growing exponentially. The data is also increasingly heterogeneous in form as text, images, and other symbolic formats. The data is also extremely diverse in context. For instance, it ranges from global guidelines based on the broadest epidemiological studies, through knowledge gained from disease-specific scientific studies, both in vitro and in vivo, to individual patient-specific data. In addition, the data also scales from the molecular to the cellular sizes, and from tissue or organ to patient representations. The recent breakthroughs in the description of the human genome and in our understanding of its connection to disease processes have also contributed more data through functional genomics studies.

This phenomenal volume of fragmented information and its growth rate represent an unprecedented data-management challenge. In particular, it is often impossible for an individual—whether a clinician responsible for patient management, or a physicist or engineer developing new imaging or interventional devices—to understand and assimilate this knowledge. The result is that it has become increasingly evident that new methods are required to manage, integrate and search data so that it becomes accessible to the end user. The @neurIST project was designed to address this issue.¹

The current fragmentation of relevant data effectively compromises disease treatment. This is why the @neurIST project first aims to achieve vertical integration across data structures and across length scales. It also seeks to take into account the horizontal integration that is present at every level of abstraction, from access to information sources, evidence processing, knowledge representation, structuring, and fusion. While the project is only focused on one carefully selected disease, cerebral aneurysm (see Figure 1), our overall goal is to create an integrative approach, scalable and reusable for other disease processes.

Figure 1. @neurIST subject-specific computational fluid dynamics (CFD) models of cerebral aneurysms are used to understand the role of blood flow in the rupture process. The pictures show simulations (color-coded by flow velocity magnitude) in four subjects (1-4) with mirror aneurysms in the Circle of Willis (i.e. aneurysms symmetrically placed in the cerebral circulation). In each subject, the aneurysms are comparable in shape, type and location, with identical genetic and systemic factors. However, one of them ruptures (a) while the other does not (b). The complex processing chain developed in the project will help to understand whether flow conditions differ and how they can be used to define surrogates of risk of rupture in a patient-specific context. Images are a courtesy of Alessandro Radaelli from the University of Pompeu Fabra and were produced in collaboration with Neuroangiografia Terapeutica SE (NAT/HGC), the Hospital Clinic i Provincial de Barcelona (HCPB) and George Mason University, using the ANSYS CFD solver.

Although important due to its impact on public health, cerebral aneurysm also has a number of interesting challenges that make it attractive as a proof of concept for our data management approach. Additionally, we believe that such a level of focus is necessary to credibly address the expected vertical integration and to identify clear exploitation paths. The latter are operative in industrial contexts, for instance in decision-support systems and in the advanced design of medical devices, and also in medical contexts, where they support further research and discovery, such as linking the molecular level understanding of a disease with the disease process itself.

The current understanding of cerebrovascular aneurysm disease combined with modern imaging technology can increasingly reveal silent lesions of the type found during clinical exams. With a prevalence of 2-3% and a rupture risk of the order of 1/10,000 people annually, treatment of this disease would benefit significantly from an integrated approach and the development of a personalized risk-assessment strategy. The integration of multiple clinical data and the use of grid-based information technology (IT) systems may provide the required platform. This will reduce health care costs by optimally targeting the relevant patient population, thus avoiding unnecessary and potentially risky interventions while improving methods for minimally invasive treatment.

Currently, the @neurIST project involves work at several different levels. The overall goal is to develop a novel IT-enabled system for cerebral aneurysm management.² This requires identifying and collecting all publicly-available, relevant, and strategically important data from scientific studies.^3–5 It also entails delivering a rich, multiscale information-processing chain to provide new diagnostic indexes and insights into the process of aneurysm formation and rupture.^6–13 Another work area is the development of a set of scalable and reusable integrative suites and the demonstration of their value for understanding and managing the disease.¹ Finally, ongoing efforts are directed at providing an information and communication technology system for developing, integrating and sharing relevant biomedical knowledge as required by the integrative suites.

The @neurIST infrastructure will not only support computationally demanding tasks, such as complex modelling and simulation, but will also enable access to public and protected health databases all over the world.¹⁴ This capacity will no doubt promote the development of corresponding systems for other disease processes by demonstrating the personal and economic impact of IT-enabled information integration in the context of cerebral aneurysm management.

Although this integrated project is primarily concerned with the vertical integration of biomedical data, many of the developed concepts have obvious implications and applications within the framework of the STEP (A Strategy Towards the Europhysiome) roadmap initiative,¹⁵ funded as a concerted action with @neurIST. In particular, it provides a platform for demonstrating the benefits and viability of the ‘in silico’ model of a human being, by merging a top-down approach starting from models of body parts and organs with a bottom-up approach that models molecular interactions, pathways and cells. We believe that @neurIST is a pragmatic and concrete example of the virtual physiological human¹⁶ concept as well as a transitional project of tremendous significance for the biomedical community.