This research is concerned with the development of algorithms to model, predict and classify physiological events and their severity in intensive care units (ICUs). Currently the modern clinical environments provide clinicians with a vast amount of physiological data that are collected or measured from different sources with different resolutions. These data are intended to provide detailed information of every patient's physiological state. However, such large and disparate collections of data impose a very complicated challenge to the clinicians for integration and interpretation. As a result it is critical to develop a computer based decision making algorithm capable of providing accurate and timely decision support.

Despite considerable advancements in the field of intelligent patient monitoring, most of the current challenges and problems are yet to be solved. The major problems that need to be addressed in any clinical decision making algorithm are: heterogeneous data, large data-sets and patient specific implementation. In this dissertation we will show how models and data driven methods can be used to address these problems.

In collaboration with Children’s Hospital of Philadelphia we are exploring several diagnostics problems using techniques that involve nonlinear physical modeling as well as computational intelligence algorithms such as machine learning. A pressing problem we are investigating is prediction of Periventricular Leukomalacia, a form of neurological damage that occurs in neonates; we are also investigating optimization of CPR, especially for pediatric patients.

The following specific projects are investigated:

•   Medical Diagnostics Using Nonlinear Dynamical Analysis
•   VCADS responds to Covid-19
•   Cardioplumonary Resuscitation (CPR) Modeling and Optimization
•   Periventricular Leukomalacia (PVL) Prediction