| Title |
High-performance ECG Data Analysis System with CNN Block Structure for Cardiovascular Condition Identification |
| Authors |
고경남(Kyeong-Nam Ko) ; 강문식(Moon-Sik Kang) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.6.39 |
| Keywords |
CNN block structure; ECG data analysis scheme; ResNeXt; MIT-BIH database; F1-score |
| Abstract |
In this paper, we propose a high-performance ECG data analysis scheme to identify cardiovascular-related health conditions by applying CNN (Convolutional Neural Network) block structure. The proposed system is designed with ResNet and ResNeXt as backbones, and learning is carried out using data from the MIT-BIH arrhythmia database. By using the proposed analysis system, it is possible to identify five cardiovascular-related conditions, and this system is implemented using the Pytorch framework. In order to evaluate the performance of the implemented system, an experiment is performed using data selected from the MIT-BIH database. The classification accuracy of the proposed system trained using the evaluation data is 98.803%, which shows the best classification accuracy compared to the existing ResNet and ResNeXt models. In addition, as a result of the analysis, the F1-score is 0.9939 in the case of the excellent performance of the proposed system. |