The Research Of Assistant Diagnosis Model For Arrhythmia Based On Anifs

The most common way to diagnose arrhythmia is the ECG signal analysis. With the characteristics of practicality, high-efficiency, non-invasiveness, safety, accuracy and reproducibility, ambulatory electrocardiography (AECG, Holter) is widely used in clinical diagnosis and medical research. However, the traditional manual diagnostic method based on ECG is limited by many aspects. Take the following for example, it consumes time for its huge amount of ECG data; misdiagnosis or missed diagnosis may happen between different doctors with different professional knowledge and clinical experience. These make ECG computer-aided diagnosis necessary. With the development of computer technology and artificial intelligence technology, ECG assistant diagnosis is becoming a research hotspot. The research can benefit the early diagnosis of heart disease. Then it is very useful for saving people's lives, improving quality of patient's life, and reducing social and economic burden caused by arrhythmia.This study aimed to build an assistant diagnosis model for arrhythmia and achieve heartbeat automatic analysis. Firstly, the preprocessing filtration was taken and 9 morphological descriptors were extracted from QRS complex and VCG plane. Secondly, the training strategy and diagnosis strategy using adaptive neuro-fuzzy inference system (ANFIS) based on multi-model were made to realize heartbeat automatic classification. Then, the sensitivity and specificity of this approach were counted by using different training datasets. Finally, the comparisons with the algorithms of Kth nearest neighbour rule (Knn) and linear discriminant analysis (LDA) were also made.This paper applied ANFIS to automatic heartbeat classification for the first time. ANFIS can implement automatic rules generation and adjust parameters of membership function, having the advantage of both neural network and fuzzy inference. Meanwhile, a training model and a diagnostic model were constructed based on ANFIS. This method could overcome the limitation that a multiple-input single-output system is hard to be applied in complex classification, facilitating the training process and improving the accuracy of diagnostic systems. The comparisons with the algorithms of Kth nearest neighbour rule (Knn) and linear discriminant analysis (LDA) were also made. The statistics showed that this model could save operation time and apply in arrhythmia diagnosis.This study also analyzed the sensitivity and specificity of this approach by using different training datasets. Due to the small sample size and biased data in general training set 1 (GLS1), the sensitivity and specificity trained by GLS1 were low. However, increasing training data and adding testing data into training datasets could increase the sensitivity and specificity.In short, we built an assistant diagnosis model for arrhythmia based on ANIFS. The experimental results showed that this model could be used in automatic heartbeat classification and it could be applied to clinical application. This may provide methods for further researches on more complex arrhythmia diagnosis model.