| Personal health and many aspects of a disease condition can be monitored pervasively in real-time by using low-cost sensors and low-power wireless systems, enabling individuals to be aware of impending health risks. This pervasive health monitoring capability can potentially provide an invaluable opportunity with the focus on early detection of medical condition, followed by individually optimized treatment recommendations that may forestall or prevent serious disease. While progress in this area is underway in sensor technology, mobile computing platforms, and data transport, barriers to large scale application remain ahead. These barriers arise due to the need for a large number of sensors to effectively diagnose patient disease conditions, which generally require data fusion from a diverse set of sensors and instruments that are applied at proper times according to environment, patient behavior, and inferred patient condition. Since the required sensors may be numerous, and may not be worn comfortably and practicably at all times, a solution is required for the systematic selection of sensors at the point of use.;We present the Medical Embedded Device for Individualized Care (MEDIC), an end-to-end system architecture enabling real-time disease diagnosis, supporting local sensing and signal processing, autonomous decision support, and remote reconfiguration and control of sensing components. We implemented this system using standard, ubiquitous wireless platforms. The MEDIC architecture incorporates a new hierarchical approach to active sensor fusion that dynamically controls sensing resources to attain high inference accuracy with an optimized sensor set. This new approach, the Incremental Diagnosis Method (IDM), is a real-time embedded decision support system based on Bayesian statistics and decision analysis theory. IDM automatically selects or deselects available sensors so that the diagnostic certainty of patient condition best improved while the set of sensors used on the patient body is minimized. We present detailed evaluations of the MEDIC system in three biomedical applications: (1) Patient Gait Analysis, (2) the SmartCane System, and (3) Exercise Monitoring. Finally, this thesis also discusses the many new opportunities provided by the MEDIC/IDM system and related future research introduced by this system. |
