Bayesian Networks for Cardiovascular Monitoring

Bayesian Networks provide a flexible way of in- corporating different types of information into a single proba- bilistic model. In a medical setting, one can use these networks to create a patient model that incorporates lab test results, clinician observations, vital signs, and other forms of patient data. In this paper, we explore a simple Bayesian Network model of the cardiovascular system and evaluate its ability to predict unobservable variables using both real and simulated patient data. I. INTRODUCTION Physicians have access to many types of information when treating patients. For example, they can examine real-time waveform data like blood pressure and ECG recordings; data trends like time-averaged heart rate; intermittent mea- surements like temperature and lab results; and qualitative observations like reported dizziness, nausea, or skin color. Within the Intensive Care Unit (ICU), physicians attempt to consider as much of the relevant information as possible, but the astronomically large amounts of data collected make it impossible to consider all available information within a reasonable amount of time. In addition, not all of the data collected is helpful in its raw form, but sufficient statistics taken from such data might help physicians gain a more thorough understanding of recent changes in the patient's state. Because of this, we are exploring ways to integrate different types of patient data into more synthesized forms (see http://mimic.mit.edu/).