Human cardiovascular system identification and application using a hybrid method of auto-regression and neuro-fuzzy inference systems

By far the most popular technique for mathematical identification of the cardiovascular system is auto-regressive moving average model based on linear assumption. However, the cardiovascular system regulated by autonomic nervous functions is a complex, unclear, and non-linear system. An effective hybrid approach, which has parallel modular structure of auto-regressive and neuro-fuzzy inference system, is proposed to identify cardiovascular linear mechanism of heart function and non-linear autonomic nervous regulation functions respectively. Auto-regressive is an efficient method to identify the stationary time series. Fuzzy set theory is very suitable to systems with uncertainties and expert knowledge. Neuro-fuzzy inference paradigm imports the auto-learning property into fuzzy logic engine, therefore extracts some knowledge from data automatically. Properties of this novel approach are evaluated based on three subjects' clinic data in term of accuracy, robustness, physiological meaning, and potential implementation in clinical problems. Modified pulse wave contour cardiac output measurement is introduced being an example of real world implementation.