基于LSTM的轴流血泵仿生搏动研究

微型轴流血泵是目前临床最常使用的心室辅助装置之一,针对其恒流辅助不符合自然心脏运作的特点,进行了轴流血泵仿生装置的搭建.针对其检测、控制延迟的缺点,利用了长短期记忆神经网络LSTM预测搏动周期来提高装置的同步搏动准确性.最终得到的训练集预测与实际结果的均方根误差为8.29,而通过测试集得到的预测与实际结果的均方根误差为5.33.最后,进行了结论和误差分析,研究结果证明了轴流泵在同步仿生搏动性上的可行性.

Research on bionic pulsation of axial flow blood pump based on LSTM

The micro axial blood pump is one of the most commonly used ventricular assist devices in clinical practice. The axial flow blood pump bionic pulsation device was built to solve the problem that the constant flow assistance does not meet the characteristics of the natural heart. In addition, the long short term memory (LSTM) neural network is used to predict the beating period to improve the synchronization accuracy of the device, aim at overcoming the shortcoming of detection delay and control delay in micro axial blood pump. The root mean square error of the prediction models on the training set and test set is 8.29 and 5.33, respectively. Finally, the conclusion and error analysis are carried out, and the feasibility of the axial flow pump in the synchronous bionic pulsatility is proved.