This Small Business Innovation Research Phase II project from NeuroDimension Incorporated will develop new, biologically inspired solutions to the problem of ventilator control of human subjects. The problem is difficult because the patient is analogous to a time-varying, nonlinear plant. Adaptive inverse control is powerful enough to adapt to changing conditions while maintaining system stability. In Phase I, NeuroDimension developed a control architecture and development environment for neural inverse control and applied it to controlling a ventilator. This solution outperformed the state-of-art commercial ventilator. The goal is to develop a system that will optimize most of the major settings currently set by clinicians on present-day ICU ventilators. The complexity of these ventilators subject their operators to speculative and empirical interpretation of many ventilatory parameters, leading to potentially hazardous misuse. NeuroDimension proposed system will model and extract relevant physiologic conditions in the patient and use this information to either control the ventilator automatically or to advise the clinician on how to change the settings. The primary technical approach utilizes neural fuzzy hybrid systems, Kohonen self-organizing maps (SOMs)-a SOM clusters the input space and assigns a different model to each cluster--and switching multiple inverse controllers. The firm has assembled a unique team of experts in the fields of neural control and ventilation to accomplish this task. NeuroDimension proffers technology that has application to a number of possible products including, inverse neural control application software; an ultra-intelligent respiratory monitor; and an easy to use and optimal closed-loop ventilator controller.
