Remote assessment of circulation is expected to be valuable in protecting individuals in the community predisposed to circulatory dysfunction, individuals involved in hazardous duties, and traditional hospital patients. This task requires sensors on the body, which ideally will be worn for days at a time during a diversity of activities, yielding real time data. Several modalities, such as photoplethysmography and electrical bioimpedance, satisfy the criteria for truly wearable biosensors; they are unobtrusive, non-invasive, and power efficient. However, the data currently derived from these sensors are not adequate for circulatory monitoring during activities of daily living, or more demanding activities such as fitness training or military battle. There are four key shortcomings in practice today: each measured signal is (a) altered by the location in which the measurement is made; (b) determined by an uncertain combination of central hemodynamic and local effects; (c) lacking in complete information necessary to characterize the central state; and (d) corrupted by local sources of noise such as physical movement and muscular bioelectrical signals. To overcome these problems, we propose a new approach to multiple sensor fusion: the multichannel blind system identification (MBSI) technique. MBSI allows the estimation of both an unknown input and the unknown dynamics from a set of related outputs. The technique is now used in wireless communication systems in which a single broadcast signal is altered by transmission across different paths then received simultaneously by multiple receivers. Similarly, it is proposed that the signals from multiple circulatory biosensors may be processed with MBSI algorithms. The MBSI algorithms may provide a systematic way of fusing multiple non-invasive biosensors in real-time to determine the cardiac output and other global hemodynamic parameters, as well as to characterize the vascular dynamics that are unique to each sensor channel. Milestones for this project will include the design of wearable, unobtrusive, low-power, non-invasive sensor hardware and MBSI data-fusion algorithms capable of 1) long-term cardiac output estimation, 2) continuous arterial blood pressure monitoring, and 3) detection of focal peripheral vascular pathologies. ? ?
McCombie, Devin B; Reisner, Andrew T; Asada, Haruhiko Harry (2005) Laguerre-model blind system identification: cardiovascular dynamics estimated from multiple peripheral circulatory signals. IEEE Trans Biomed Eng 52:1889-901 |