Mobile Menu

AI-based approach for detecting myocardial scarring

Researchers have developed an artificial intelligence (AI)-based approach for the prediction of myocardial scarring based on electrocardiogram (ECG) and additional clinical parameters.

Ischaemic heart disease

Ischaemic heart disease (IHD) is a major cause of death in Westernised countries. It has both a personal and high socioeconomic impact. Early detection of IHD can enable rapid treatment initiation and in turn can reduce mortality. Myocardial scarring occurs during the preliminary stage of IHD and therefore, acts as an early indicator of disease. Consequently, early detection of myocardial scarring, alongside appropriate therapeutic measures, could prevent IHD or reduce the effects of the disease.

MRI is the gold standard for diagnosing myocardial scarring. However, there are several drawbacks of the application of MRI as a screening method, including limited availability, high costs and its time-consuming nature. Myocardial scarring can impact ECG signals, yet the effects are complex to interpret and do not achieve required diagnostic sensitivity.

Several studies have explored the use of machine learning or deep learning algorithms for ECG-based diagnosis of cardiac diseases. For automated detection of myocardial scarring, so far only classic machine learning support vector machines (SVMs) have been used. This method requires pre-extraction of features to make ECG recording accessible for processing.

Detection of myocardial scarring

In this study, published in Biological Chemistry, researchers developed a deep learning model for the detection of myocardial scarring through raw ECG recordings and corresponding clinical parameters. Researchers trained and evaluated the model by applying 6-fold cross-validation to a dataset of 12-lead ECG time series, together with clinical parameters.

The team found that the model achieved a sensitivity of 70.0%, specificity of 84.3% and accuracy of 78.0%. Overall, the developed deep learning model reached promising performance measures to detect scarring. This model could be applied directly to ECG data and clinical parameters and does not require pre-extraction features. The team believe that this high diagnostic precision for myocardial scarring detection may support a novel, comprehensive screening method. Additionally, this model could be used to potentially capture phenotype during clinical trial studies or to identify specific cohorts for further study.

Image credit: By –

More on these topics

AI / Heart Disease / Machine Learning

Share this article