SlideShare utilise les cookies pour améliorer les fonctionnalités et les performances, et également pour vous montrer des publicités pertinentes. Si vous continuez à naviguer sur ce site, vous acceptez l’utilisation de cookies. Consultez nos Conditions d’utilisation et notre Politique de confidentialité.
SlideShare utilise les cookies pour améliorer les fonctionnalités et les performances, et également pour vous montrer des publicités pertinentes. Si vous continuez à naviguer sur ce site, vous acceptez l’utilisation de cookies. Consultez notre Politique de confidentialité et nos Conditions d’utilisation pour en savoir plus.
International system for total early disease detection (INSTEDD) platform
International System for Total Early Disease Detection (InSTEDD) Platform Taha A. Kass-Hout, M.D., M.S., Nicolas di Tada InSTEDD, Palo Alto, California OBJECTIVE tected in location X, and with a certain spatio-This paper describes a hybrid (event-based and indi- temporal pattern”). The human input and reviewcator-based) surveillance platform designed to module is exposed as a set of functionalities that al-streamline the collaboration between domain experts lows users to comment, tag, and rank the elementsand machine learning algorithms for detection, pre- (positive, neutral, or negative). Additionally, usersdiction and response to health-related events (such as can generate and test multiple hypotheses in parallel,disease outbreaks). further collect and rank sets of related items (evi- dence), and model against baseline information (for BACKGROUND cyclical or known events). The platform maintains aOver the last decade, the majority of the designs, list of ongoing possible threats allowing domain ex-analyses and evaluations of early detection  (or perts to focus their field information and either con-biosurveillance) systems have been geared towards firm or reject the hypotheses created. That feedbackspecific data sources and detection algorithms. Much is then fed into the system to update (increase or de-less effort has been focused on how these systems crease) the reliability of the sources and credibility ofwill "interact" with humans . For example, con- the users in light of their inferences or decisions.sider multiple domain experts working at different RESULTSlevels across different organizations in an environ-ment where numerous biosurveillance algorithms The platform synthesizes health-related event indica-may provide contradictory interpretations of ongoing tors from a wide variety of information sourcesevents. This paper discusses the anticipated contribu- (structured and unstructured) into a consolidated pic-tion of social networking, machine learning and col- ture for analysis, maintenance of “community-widelaboration techniques to address these emerging is- coherence” , and collaboration processes. Thissues by drawing upon methods, models and tech- helps detect anomalies, visualize clusters of potentialnologies that have been proven to work in other chal- events, predict the rate and spread of a disease out-lenging domains. break and provide decision makers with tools, meth- odologies and processes to investigate the event. METHODS Presently, the platform and associated modules areThe platform consists of several high-level modules, being piloted for the Mekong Basin region in SEincluding: 1) Data gathering, 2) Automatic feature Asia.extraction, data classification and tagging, 3) Human CONCLUSIONSinput, hypotheses generation and review, 4) Predic-tions and alerts output, and 5) Field confirmation and In this paper we describe a platform that enables de-feedback. The data gathering module allows users tection, prediction and response to health-relatedto collect information from several sources (SMS events through a collaborative approach that com-messages, RSS feeds, email list (e.g., ProMed), bines data exploration, integration, search and infer-documents, web pages, electronic medical records, encing – providing more complex analysis andanimal disease data, environmental feed, remote deeper insight. We believe that such a platform repre-sensing, etc.). The automatic feature extraction, sents the next generation of early detection and re-data classification and tagging module is an exten- sponse systems.sible architecture that allows the introduction of ma- REFERENCESchine learning algorithms (e.g., Bayesian). These  Ping Yan, Daniel Dajun Zeng, Hsinchun Chen: A Review ofcomponents extract and augment the features (or Public Health Syndromic Surveillance Systems. ISI 2006; 249-metadata) from multiple data streams; such as: source 260.and target geo-location, time, route of transmission  Thagard, P. Coherence, truth, and the development of scientific(e.g., person-to-person, waterborne), etc. In addition, knowledge. Philosophy of Science 2007;74, 28-47.these components help detect relationships between  Billman, D., Convertino, G. Shrager, J. Massar, JP. and Pirollithese extracted features within a collaborative space P. The CACHE Study: Supporting Collaborative Intelligence.or across different collaborative spaces. Furthermore, Paper presented at the HCIC 2006 Winter Workshop on Collabora- tion, Cooperation, Coordination. Feb. 2006.with human input, these components can suggestpossible events or event types (e.g., at the earliest Further Information:stages of a disease outbreak: “there is an unknown Taha Kass-Hout, firstname.lastname@example.org event, transmitted person-to-person, de- http://www.instedd.org Advances in Disease Surveillance 2008;5:108