抄録
To simulate a biological relationship between arterial pressure (AP), renal sympathetic nerve activity (RSNA), heart rate (HR) and respiration rate (RR), we have built a new simulation model by a neural network (NN) algorithm (Neural Network Toolbox of MATLAB) using only measurements but without approximation equations. The learning algorithm of the layered NN was a back propagation. AP, RSNA, HR and RR were measured in conscious chronically instrumented rats. Previously, averaged values of experimental observations were used for the learning of NN model, while in this study, individual values of experimental observations were used for the leaning. The errors generated from the present simulation were smaller than those generated from the previous method. Using the present simulation model that was constructed using the data from normotensive Sprague Dawley (SD) rats, we examined whether the simulated AP values would be consistent with the measured AP values in salt-sensitive hypertensive Dahl rats, or chronic heart failure Dahl rats. The simulated AP values were significantly lower than the measured values. The results suggested that the NN modeling could indicate that the biological relationship between RSNA, HR, RR and AP in hypertensive or heart failure state was theoretically different from that of the SD rat strain. [J Physiol Sci. 2007;57 Suppl:S210]
