Models, systems, networks in economics, engineering, nature and society
Background. The article is devoted to the development of a neural network for ECG signals classification. Automatic classification of ECG signals frees cardiologists from laborious and monotonous work and reduces the time of ECG interpretation. The aim of the study is to create and evaluate a convol...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Penza State University Publishing House
2025-02-01
|
| Series: | Модели, системы, сети в экономике, технике, природе и обществе |
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Background. The article is devoted to the development of a neural network for ECG signals classification. Automatic classification of ECG signals frees cardiologists from laborious and monotonous work and reduces the time of ECG interpretation. The aim of the study is to create and evaluate a convolutional neural network model for automatic ECG signals classification in 12 standard leads to identify the most common and dangerous cardiovascular diseases. Materials and methods. Groups of diseases for classification were selected and substantiated. An original, modified architecture of the 1D ResNet34 convolutional neural network was proposed. ECG recordings from the publicly available Chinese CPSC Database were used to train and test the model. The training and evaluation of the model's performance were carried out using the 10-fold cross-validation method. Results. The performance of ECG signal classification was evaluated using standard metrics. The average values of accuracy, F1 score, and AUC-ROC for the developed classifier are 0.964, 0.832, and 0.975, respectively. Conclusions. The performance of the model corresponds to the world level of the best global achievements and is comparable to the expert-medical level. The developed ECG signal classifier can be integrated into various electrocardiographic diagnostic systems. |
|---|---|
| ISSN: | 2227-8486 |