MI Identification Algorithm Based on Feature Selection via Chi-square Test

Abstract

The commonly used myocardial infarction (MI) detection data in medicine is the electrocardiogram (ECG) signal, but it has a large feature dimension and contains a lot of noise, non-informative features, and weakly correlated features. To address this issue, this paper proposes an MI identification algorithm based on feature selection via chi-square test. Compared with other recognition algorithms, this paper uses filter and chi-square test to reduce noise and weakly correlated features in the ECG signal and achieves dimensionality reduction and visualization. To validate this identification algorithm, 12 lead ECG data from patients with MI and normal patients are obtained from the PTB diagnostic ECG database. The experimental results show that the proposed algorithm recognition rate of 99.05% achieves satisfactory results.