ICU ventilator modeling via neuro-fuzzy system (R-POPTVR)

Abstract

In an Intensive Care Unit, artificial ventilation is a key component in supporting life. However as medical technologies become increasing advanced, the rapidity and complexity of changes in ventilator machine control becomes much of a challenge where unfamiliar jargon and technical detail render it confusing and formidably for less experience physicians and clinicians to manually control them. Studies had shows that a large percentage of ventilators related deaths and injuries are caused by human error. Hence, there is a need for an expert system to assist in the controlling of the ventilator to ensure proper administration of oxygen to patients. This project will examine the possibility of modelling a medical ventilator with an online R-POPTVR neural network. To resolve the drawback of the POP rule identification algorithm where it has to consider all the possible rules at the beginning of the learning process, another rule identification algorithm call the LazyPOP will be implemented in the online R-POPTVR system. In addition, a self organization Gaussian Discrete Incremental Clustering (gDIC) technique is implemented in the online system to automatically form fuzzy sets in the fuzzification phrase. This clustering technique does not require having prior knowledge about the number of clusters.