EcgHub: in-depth analysis of digital ECG
EcgHub is a 12-lead ECG labeling system and neural network models’ collection for pathology prediction. The system allows to predict the presence or absence of several pathologies, as well as to perform and review the syndromic ECG markup based on the verified questionnaire, thus providing a dataset for further development of neural network models.
Features and advantages
EcgHub is based on research results in the areas of digital signal processing and machine learning algorithms. The pathology classification system is based on deep neural networks. The expert-verified approach provides consistent ECG labeling for training and further development of predictive models for screening and diagnosis of cardiovascular diseases.
- Trained neural networks for pathology prediction of digital ECG;
- Continuous development and refinement of neural network models, including fine-tuning for relatively small medical datasets;
- Adaptation of trained neural network models for pathology prediction of single-lead ECGs (cardiac chair, smart watches) as well as 24-hour ECGs (Holter monitoring);
- A consistent syndromic markup system to provide qualitative data for training predictive models;
- Integration of neural network models into the customer's digital circuit or remote access to the service at ISP RAS;
- Applying the markup system in the education of modern functional diagnostics professionals;
- Development of an automated population screening system.
Who is EcgHub target audience?
- Medical institutions: the prediction of neural network models can be used as a second opinion;
- Educational institutions: verified datasets allow evaluating the knowledge of students or novice doctors of relevant specialties;
- Developers of devices and applications that perform ECG diagnostics autonomously.
EcgHub deployment stories
The neural network model of classification of 12-channel ECGs is trained on the data of the Republic of Tatarstan, integrated in the pilot operation mode into the "Unified Cardiologist" system and tested on ECG data from different regions (the Republic of Tatarstan, Moscow, Veliky Novgorod).