Application of convolutional neural network in automatic classification of arrhythmia

Jian Liu,Mingxuan Fu,Shasha Zhang
DOI: https://doi.org/10.1145/3321408.3326660
2019-01-01
Abstract:As a typical disease, heart disease seriously threatens human health, and its suddenness has hindered the diagnosis and treatment for the patients. Therefore, a fast and accurate real-time analysis method is useful to assist doctors in disease diagnosis. In this paper, we mainly design three different automatic classification algorithms to classify the heart beats of electroencephalography. The algorithm is mainly based on one-dimensional convolutional neural network (CNN), and the appropriate network structure is found by comparing the classification performance of different structural models. We also use the support vector machine (SVM) and stacking methods to improve the CNN network. The data used in this paper are all from the MIT-BIH Arrhythmia Database. We use the ten-core cross-validation method to classify the four types of heart beats. The results show that among the three types of algorithms, the stacking algorithm works best and the classification accuracy is as high as 99.1%.
What problem does this paper attempt to address?