用连续量输入的人工神经网络评定心脏收缩功能

Evaluation of human heart systolic function by neural network withanalogue input

作者应用人工神经网络技术，实现心脏收缩功能的计算机自动评定，以探讨人工神经网络技术在生物医学评定、分类和诊断等问题中的应用方法，该研究采用心脏收缩时间间期的７项指标，应用反向传播神经网络，以连续量作为网络输入，对２０２位受试者的心脏收缩功能进行评定，人工神经网络的评定结果与专家评定结果相比较，符合率可达９５．５４％．结果表明，连续量输入方式在逼近实际分类边界，减小网络规模，加快网络学习方面优于数字量输入方式．

In the present study, we assessed heart systolic function using back-propagation neural networks.Applications of neural networks in the classification,diagnosis and evaluation of biomedical problems werediscussed.Heart systolic function of 202 subjects was assessed using seven indexes of systolic time intervals.The results demonstrated an accordance of 95.54% as compared to assessment by experts. In the study,sev-eral speeding methods were employed to improve the characteristics of the neural networks and their applica-tions indicated that they may be very valid for the neural networks. The findings of our study indicate thatthe artificial neural networks provide a new method for the classification,diagnosis and evaluation of medicalproblems.