The Aide Diagnosis of Cardiac Heart Diseases Using a Deoxyribonucleic Acid Based Backpropagation Neural Network

In this paper, a novel diagnostic scheme of cardiac heart diseases is presented using a Deoxyribonucleic Acid (DNA)-based BP(DNA-BP) Neural Network by distinguishing the shapes of ST segments automatically. First, wavelet transform is applied to extract the ST segments by identifying the characteristic points in the ECG. ECG signals are decomposed by aTrous Algorithm using dyadic spline wavelets. The relationship between the feature points of ECG signals and the modulus maximum pairs of the signals' wavelet transform is established, and the R-wave and ST segment's fiducial points are extracted at different wavelet scales. Second, in order to overcome the disadvantages of the BP neural network, a DNA optimization method is adopted to optimize the original weights and bias of a BP neural network. The BP algorithm is used to find the most optimal values of the weights and bias of the BP network. At last the effectiveness of the proposed aided diagnostic scheme is demonstrated via the experiments which the data are from the clinical study and MIT/BIH ECG data base. In order to validate the advantages of the DNA-based BP (DNA-BP) network and determine the best application conditions of the network, two types of experiments were conducted. In each experiment, 30 samples were selected as the training data and testing data respectively. In the first type of experiments, the data were obtained from the ECG of different people. While in the second type of experiments, the data were obtained from the ECG of one person at different times. The same data and the conditions were used in both BP and GA-based BP (GA-BP) networks to illustrate the advantages of the proposed DNA-based BP network. The experiment results illustrate that the proposed DNA-based BP network overcomes the limitation of the sloping method in identifying straight line ST segments and the limitation of function fitting method in fitting accuracy. It also surmounts the disadvantages of a BP network in terms of the local minimum, slow convergence, and the shortcomings of GA-BP in terms of its limitation in algorithm coding and evolution ways. In conclusion, the neural network is used to supply an invaluable aid for physicians to diagnose cardiac heart diseases and myocardial infarction.