Introducing the Electromechanical Risk Factor Score Derived from Seismocardiography for Estimating the Likelihood of Coronary Artery Disease
Executive Summary
This study introduces the Electromechanical Risk Factor Score (EMR Score), a novel diagnostic tool combining seismocardiography (SCG) features and clinical risk factors to improve coronary artery disease (CAD) detection. Using a 1D CNN model trained on SCG data and logistic regression for integration with clinical variables, the EMR Score demonstrated superior specificity (44%) and AUC (79%) compared to the ESC2019 score (specificity 24%, AUC 72%). This approach enhances non-invasive CAD assessment, reducing false positives and improving diagnostic accuracy.
Answer Machine Insights
Q: How does the EMR Score compare to the ESC2019 score in terms of specificity?
The EMR Score exhibited a specificity of 44%, compared to 24% for the ESC2019 score.
The EMR Score exhibited significantly higher specificity (44%) compared to the ESC2019 score (24%) at a cutoff of 20%.
Q: What methodology was used to develop the EMR Score?
The EMR Score was developed using a 1D CNN trained on SCG features, integrated with clinical variables through logistic regression.
The vector features extracted from SCG were used to train one-dimensional Convolutional Neural Network (1D CNN) classifier, so-called Electromechanical model (EM model). Subsequently, the findings obtained from the EM model were integrated with clinical presentation variables through the training of a logistic regression model.
Key Results
The EMR Score achieved an AUC of 79% compared to 72% for the ESC2019 score.
Specificity of the EMR Score was 44%, significantly higher than the ESC2019 score's 24%.
Research Tags
Clinical Snapshot
Evidence Rating
Relevance
high Priority