Abstract
We simulated the dual activity sensors rate adaptive pacer with neural network rate regulatiom algorithm on the computer. Two activity sensed pacers, BilogTM and SwingTM with different mechanisms of activity sensing, were selected as sencors including rate regulation algorithm. Back-propagation was chosen as a neural network paradign using NeuralWorks Professional II/PlusTM (NeuralWare, Inc.) on Macintosh IICi with system 7.01. Sixteen healthy volunteer with pacers on the left precordial region performed 2 kinds of multiple exercise circuit tests. The pulse rates of both pacers and sinus rate were recorded. For learning and recall, the sensor data (on/off) and the pulse rates of 5, 15 sec before were input in the network. The sinus rate as the requesting output was programmed for learning. The recalled data were compared with the pulse rates of both pacers and sinus rate, using the root mean square of the errors. The errors of the recalled data were 0.0396, Bilog 0.0501, and Swing 0.0630. We concluded that combining two different activity sensors with neural network rate regulation algorithm might improve the rate adaptation provision, and neural network might be useful tool for creating dual sensors rate regulation algorithm.
