Key result
The proposed Shock Advice Algorithm using a K-nearest neighbors model and 8 principal component features achieved a diagnostic accuracy of 99.52%, sensitivity of 97.69%, and specificity of 99.91% for detecting sudden cardiac arrest.
Why the study?
Sudden cardiac arrest consisting of shockable rhythms is life-threatening and treated efficiently by automated external defibrillators, prompting the development of improved shock advice algorithms for electrocardiogram diagnosis.
Does a novel shock advice algorithm using a KNN model and PCA-transformed features accurately diagnose sudden cardiac arrest from ECG signals?
Does a novel shock advice algorithm using a KNN model and PCA-transformed features accurately diagnose sudden cardiac arrest from ECG signals?
A novel shock advice algorithm using a k-nearest neighbors model and 8 PCA-transformed features achieved >99% accuracy in detecting shockable rhythms for automated external defibrillators.
May enhance AED rhythm analysis algorithms; leaves open prospective validation before any clinical adoption.
Sudden cardiac arrest (SCA) consisting of ventricular fibrillation and ventricular tachycardia considered as shockable rhythms is a life-threatening heart disease, which is treated efficiently by the automated external defibrillator (AED). This work proposes a novel design of the SAA, which includes a k-nearest neighbors model and a subset of 8 features extracted from the ECG segments, for the SCA diagnosis on the electrocardiogram (ECG) signal. These features are addressed as the most productive subset among 31 input features based on the evaluation of the feature correlation. The recursive feature elimination algorithm combined with the Boosting model and wise-patient fivefold cross-validation method is adopted for the calculation of the average feature importance, which shows the degree of feature correlation, to construct various input feature subsets. Moreover, component feature combinations known as the representatives of the input feature subsets with an enormous level of correlation and independence are transformed from the input subsets by the principal component analysis method. The wise-patient fivefold cross-validation procedure is used for the evaluation of these component feature combinations on the validation set. The proposed SAA is certainly efficient for SCA detection with a small number of the extracted feature and relatively high diagnosis performance such as accuracy of 99.52%, sensitivity of 97.69%, and specificity of 99.91%.
No takes yet. Share an insight, caveat, or question.
Pham et al. (2023) studied Sudden cardiac arrest (ventricular fibrillation and ventricular tachycardia) (n=57). Shock Advice Algorithm (SAA) using K-nearest neighbors (KNN) and Principal Component Analysis (PCA) with 8 features vs. Other machine learning and deep learning models was evaluated on Classification accuracy for shockable vs non-shockable rhythms. The proposed Shock Advice Algorithm using a K-nearest neighbors model and 8 principal component features achieved a diagnostic accuracy of 99.52%, sensitivity of 97.69%, and specificity of 99.91% for detecting sudden cardiac arrest.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: