神经网络在叶轮式人工心脏输出流量检测中的应用

The Application of Neural Network Technique for Output Estimation of an Impeller Artificial Heart

目的 寻找一种既精确又简便的间接测量方法。方法 应用神经网络技术检测叶轮式人工心脏的输出流量。采用电机功率和转速作为神经网络输入元 ,流量作为输出元的网络结构 ,通过反复训练、测试 ,使网络掌握输入与输出之间的非线性关系 ,能够在一定的误差范围内根据输入得到相应的输出。结果 应用神经网络检测人工心脏输出流量能够达到一定的精度 (误差 <5 % )。结论 该方法简便、精确 ,在实际测量中避免引入传感器探针 ,降低装置的复杂性 ,减小了感染的机会 ,具有进一步研究价值。

Objective An accurate,convenient and indirect measurement for artificial heart was developed. Methods The neural network technique was applied to estimate the output of an impeller artificial heart. The neural network construction including two input units-motor input power, rotational speed and one output unit-rate of flow, was set up. By training and testing repeatedly, the network can accurately learn the non-linear relationship between the input units and the output units, and estimate the output value according to input values. Results The application of a neural network can solve the problem successfully with an error under 5%. Conclusion The method was accurate and convenient. Moreover, with no need of sensor, it was less complex and reduced the infection possibility. Thus, it is worth studying further.