基于人工神经网络的心脏功能评定研究

Investigation on Evaluating the Human Heart Function with Neural Network

作者应用人工神经网络的方法,利用心脏收缩时间间期指标评定心脏功能。采用21个输入、3个输出和单个隐层的前馈网络,用反向传播算法进行训练。所用7个指标,将其进行编码后作为输入矢量,心脏功能分为3级,在人工神经网络学习由专家评定的结果后,对200位受试者的心脏功能进行评定。人工神经网络评定的正确率达93.5％,且具有自学习、容量扩充和较强的容错能力。

In this paper, we discuss the application of a back-propagation neural network to the evaluation of the human heart function. Currently, this assessment is made by experts skilled in the interpretation of human heart functidn according to the cardio systolic time intervals (STI) obtained noninvasivaly. Our target is to develop a machine-based methodology for the assessment of heart function.In this research, the assessment problem of the heart function will be approached using a computing paradigm known as neural networks. The 'back-propegation' learning algorithm has been used to train feedforward network to perform the problem. There are 7 criteria used in the study. On the basis of the quantitative studies, three ordered classes(normal, doubt, abnormal)were defined. The neural network has 3 outputs with '100' indicating class 1, '010' class 2 and '001' class 3. Each criterion is converted into 3 bit-binary codes for the network input, and the feature vector is a 21-component binary vector. The number of hidden layer is one. The data set contains a total of 200 observations and is divided into training and test sets containing 60 and 140 observations, respectively. The distribution of classes in the training set is uniform. The simulation results in 93.5% percent correct assessment for the test set.