Research On Automatic Diagnostic System Of The Heart Disease Based On ANFIS And Wavelet Transform

Electrocardiograph (ECG) is a practical, efficient, noninvasive, safe, accurate and reproductive way to diagnose cardiovascular disease. That is why it is widely used in clinical diagnosis and other medical researches. But the traditional manual diagnostic method based on ECG is limited by doctors' professional knowledge and clinical experience, and it consumes time for its huge number of the ECG data. So it is necessary for us to develop an ECG automatic diagnostic system. With the development of computer technology, the study of ECG automatic diagnosis has recently received attention. The research can benefit the early diagnosis of heart disease. Then it is very useful for saving human's lives and improving the quality of people's life. So the ECG automatic diagnostic system has great economic and social significance.The main purpose of this paper is to research electrocardiogram automatic diagnostic system. Firstly, wavelet transform is used to remove the noises in ECG signal and to extract characteristic points. Then, an adaptive neuro-fuzzy inference system (ANFIS) fault diagnosis strategy is first constructed to analyze ST segments. The main contents are as follows:(1) ECG high frequency noises, such as power-line interference and myoelectricity interference, and low frequency noise, such as baseline wander, have been inhibited using haar wavelet transform. Then ECG signals are decomposed by aTrous Algorithm using dyadic spline wavelets. The relation between the characteristic points of ECG signals and the modulus maximum pairs of the signals' wavelet transform is illustrated. Taking advantage of the multiple resolution capacity of wavelet transform, a scheme is developed to identify the R-wave and ST segments characteristic points at different wavelet scales automatically. Finally, the applicability of the strategy is demonstrated through various data.(2) ST segment contains the main information of ECG, so its analysis is very important for the ECG automatic diagnosis. According to chemical research, all ECG data are classified into six basic types. Feature points of ST segment are chosen as the inputs of the diagnostic system under the doctors' guidance. Since the feature points can't fully contain information of ST segments, two fitted slopes of ST segments are as a part of the input vectors to train the constructed ANFIS diagnostic system. (3) Due to its complication of training a multi-input-multi-output ANFIS, a fault diagnostic system is constructed in this paper, which is composed of training system and diagnostic system. At the end the precision of the diagnostic system has been quantitatively analyzed.(4) Five abnormal shapes of ST segments are considered as the faults of the diagnostic system. The results show that it is successful using ANFIS strategy in diagnosing ST segment shapes. Compared with BP method, the accuracy and anti-jamming can be improved.