Abstract
A model of the cardiovascular system was developed using a neural network technique. A treadmill exercise experience was performed to obtain the physiological signals during the exercise with a goat. The blood pressure catheter and blood flow prove were implanted in the chest and the exercise tests were performed after two weeks of the operation. The model to estimate cardiac output from heart rate, mean arterial pressure, mixed venous saturation and physical activity that were obtained from the exercise experience was constructed based on a neural network technique. A three-layer back-propagation network, which has the input layer with forty two cells, the hidden layer with thirty two cells and the output layer with 10 cells, was used to construct the model. Five hundred and forty six sets of data were used for learning process of the network. Seventy sets of the learning data and sixty six sets of data which were not used for learning were used to evaluate the performance of the cardiac output estimation. The correlation coefficient between the learning data and real cardiac output was 0.965 (p<0.001) and that for the non-learning data was 0.869 (p<0.001) and the feasibility of the neural network for modeling of cardiovascular system was confirmed.
