Modelling and monitoring the individual patient in real time

This paper presents a framework for representing background knowledge and new data, and reasoning efficiently with this powerful combination of both knowledge and data. Domain knowledge is needed to positively bias the operation of data mining algorithms. Knowledge in the form of mathematical models can be considered sufficient statistics of all prior experimentation in the domain, embodying generic or abstract knowledge of it. We present a framework for using this knowledge in a probabilistic framework for data mining, inference, and decision making under uncertainty. Real-time data-streams, which typically contain uncertainty, are then exploited in a principled manner to individualise patient care. By combining the knowledge available in existing data streams with the expert knowledge available and an efficient inference method, we provide a powerful foundation for reasoning with uncertain and sparse data in the medical domain.