|Biosurveillance-based Integrated Outbreak Warning And Recognition System (BIOWARS)|
The BIOWARS technology is an adaptive system for discovering disease outbreaks and impending bio terrorism attacks. BIOWARS uses syndromic surveillance to find symptomatic data patterns, and applies Bayesian networks in collecting and archiving these patterns.
An important challenge faced by intelligence analysts and the intelligence community in our post 9/11 world is to gather, piece together, and correctly interpret vast amounts of intelligence data--data that may signal an impending attack or that may help limit the severity of an attack. As a Defense Science Board study of transnational threats noted, the “making of connections between otherwise meaningless bits of information is at the core of (transnational) threat analysis.”
This challenge is particularly significant with respect to biological emergencies: i.e., emergencies resulting from bio terrorism or a major disease outbreak. Monitoring for such emergencies involves a host of distributed and disparate information sources—intelligence sources, public health sources, open source media, etc.—that must be mined and their data compiled in a format conducive to analysis and interpretation. Current outbreak detection/threat prediction approaches rely on disease-based surveillance, which involves extensive disease testing and the use of cumbersome data systems for relaying test results. These approaches also tend to rely on systems and channels of information flow that report known, diagnosed cases--hospital information systems, emergency physician reports, school or work absenteeism, agricultural surveillance—and that are slow, can be unreliable, and make little or no provisions for the reporting and use of intuitive observations based on anomalous patterns and trends. These methods are time consuming and unresponsive, and may be ineffective in stemming the spread of rapidly communicable disease outbreaks or of biological agents such as anthrax.
In this initiative, KBSI is designing, developing, and deploying an adaptive system for the automated discovery of disease outbreaks and impending bio terrorism attacks. The main product will be the Biosurveillance-based Integrated Outbreak Warning And Recognition System (BIOWARS), an agent-based decision support system that facilitates the automated generation of information from disparate and distributed sources to support the early detection of emerging threats and potential disease outbreaks. Unlike current surveillance methods, BIOWARS applies syndromic surveillance, a knowledge based approach that looks for patterns of symptoms in the data that relate to or indicate common biological threat agents. BIOWARS uses Bayesian networks and models to collect, organize, and store knowledge concerning these data patterns, yielding meaningful information and that can be used to generate an overall threat assessment. The Bayesian network models can be managed and improved through use, enabling the tool to adapt to changing data patterns and improve its performance over time.
BIOWARS represents an innovative knowledge-based approach to biosurveillance that combines the power of data mining and knowledge discovery methods with classical rule-based expert systems techniques. BIOWARS will greatly improve on current surveillance practices by integrating the many open sources of information from which analysts gather data, organizing data into adaptive models and thus speeding threat detection and reducing false alarm rates. This technology will have immediate benefits to the military intelligence community and, in concert with a number of related KBSI technologies, benefit military and private sector health organizations involved in monitoring disease outbreaks and emergency medical response.
Phase II Development
In Phase II of the BIOWARS initiative, KBSI is developing and demonstrating a Biosurveillance-enabled BIOWARS system as part of a much larger collaborative effort to enhance our national biosurveillance capability. The final goal of the effort is to deploy this biosurveillance system. KBSI is actively pursuing semantic web capabilities for the prototype Biosurveillance Community of Interest (BCOI) website, including the development of grid processing and data distribution capabilities (using the Globus and HaDoop technologies) as well as visualization, text mining, and data mining capabilities. This effort, as a key contributor to a much larger vision, is central to the overall success of the goal of protecting the nation from biologically based threats and risks.
An important development in support of this goal is the recent decision by the Department of Health and Human Services (DHHS) and the American Donor Biovigilance community to adopt the BCOI technology for use in their community. These organizations have also made significant headway in introducing our technology into the International Society for Blood Transfusions as a mechanism for enhancing collaboration within that community. KBSI is continuing to expand the prototype BCOI portal implementation and further develop application technologies for the BCOI community.
In addition, the DHHS, in collaboration with leading U.S. blood centers and blood banks, the FDA, and the CDC, has developed a Donor Biovigilance System (DBS) that focuses on capturing the reactions encountered by donors during blood donation. The American Association of Blood Banks (AABB) and the nation’s leading blood centers, including the American Red Cross and Blood Systems Incorporated, are participating in the design and development of this system.
DBS will serve as an application test-bed for BIOWARS, providing an established community of interest that can provide requirements input and end user testing and evaluation. The challenging data mining and analysis requirements of DBS will be used to guide the BIOWARS research and will serve as a foundation for designing the analytical components of the BCOI solution. In addition, the collaboration features of BIOWARS (chat, wiki, discussion forums, tagging) will be transitioned to DBS.
Related Links & Downloads