Methods for Improved Discrimination Between Ventricular Fibrillation and Tachycardia.

Differentiating between ventricular tachycardia and ventricular fibrillation in clinical and preclinical research is based on subjective definitions that have yet to be validated using objective criteria. This is partly due to shortcomings in the discrimination ability of current objective approaches, typified by the algorithms that perform cardiac rhythm classification using low-dimensional feature representations of electrocardiogram (ECG) signals. These identify ventricular tachyarrhythmias, but do not discriminate between ventricular tachycardia and ventricular fibrillation. In order to address this limitation, we have tested the utility of high-dimensional feature vectors, in particular, magnitude spectra and classifier ensembles that take into account local context information from ECG signals. Using these approaches, we categorized rhythms into three classes: ventricular tachycardia, ventricular fibrillation, and any other possible rhythm, defined here as "nonventricular rhythms." The high-dimensional spectral features achieved a substantial improvement in the discrimination between ventricular tachycardia and ventricular fibrillation, but exhibited a decreased sensitivity to nonventricular rhythms. In order to deal with the reduced sensitivity for the detection of nonventricular rhythms, methods were elaborated for combining the strengths of different feature spaces, and this substantially improved the identification sensitivities of all three classes.

Structured prediction for differentiating between normal rhythms, ventricular tachycardia, and ventricular fibrillation in the ECG.

Recent studies have been performed on feature selection for diagnostics between non-ventricular rhythms and ventricular arrhythmias, or between non-ventricular fibrillation and ventricular fibrillation. However they did not assess classification directly between non-ventricular rhythms, ventricular tachycardia and ventricular fibrillation, which is important in both a clinical setting and preclinical drug discovery. In this study it is shown that in a direct multiclass setting, the selected features from these studies are not capable at differentiating between ventricular tachycardia and ventricular fibrillation. A high dimensional feature space, Fourier magnitude spectra, is proposed for classification, in combination with the structured prediction method conditional random fields. An improvement in overall accuracy, and sensitivity of every category under investigation is achieved.