Scalable, Multiplatform, and Autonomous ECG Processor Supported by AI for Telemedicine Center

Background: Wearable devices play an essential role in the early diagnosis of heart diseases. However, effective management of long-term $ECG$ measurements (1-3 weeks) by a telemedicine center $(TMC)$ requires specifically designed software. Method: We used the multiplatform framework.NET to build the application. Deep-learning models for $QRS$ detection, classification, and rhythm analysis were trained in the PyTorch framework; models were trained using data from Medical Data Transfer, $s. r. o$ . Czechia $(N=73,450$ and 12,111). The ONNX runtime libraries were used for model inference, including acceleration by graphic cards $\cdot$ Results: The pre-production benchmark (recordings of 82 patients) showed a mean accuracy of $0.97 \pm 0.04$ for $QRS$ detection and classification into three classes; it also showed a mean accuracy of 0.97 $\pm 0.03$ for rhythm classification into seven classes. Conclusion: The presented software is a fully automated, multiplatform, and scalable back-end application to process incoming $ECG$ data in the $TMC$ Although it is not freely accessible, we are open to considering processing $ECG$ data for research and strictly non-commercial purposes.