Preliminary results on the prediction of countershock success with fibrillation power

Fred A. Hamprecht, Daniel Jost, M Rüttimann, Franck Calamai, J KowalskiResuscitation 2001 · DOI

TL;DR

This paper explores new ways to predict whether an electrical shock will successfully restart a heart in patients experiencing ventricular fibrillation (VF). They found that "fibrillation power," a measure of the electrical chaos in the heart, performs comparably to existing methods like dominant frequency and mean amplitude, suggesting it's a promising, complementary predictor.

What Problem Does This Paper Address?

When a person's heart stops due to ventricular fibrillation (VF), doctors use electrical shocks (defibrillation) to try and restore a normal rhythm. Predicting whether a shock will be successful before delivering it is crucial for optimizing treatment and patient outcomes. Current methods for predicting success, like analyzing the dominant frequency or amplitude of the VF electrocardiogram (ECG), have limitations. This paper investigates a new descriptor, "fibrillation power," which was previously validated in animal studies, to see if it can improve prediction accuracy in human VF episodes.

Methods

The researchers analyzed ECG recordings from 54 patients who experienced out-of-hospital cardiac arrest with VF. For patients who had a return of spontaneous circulation (ROSC), they extracted ECG segments from just before the successful shock. For those who did not achieve ROSC, segments were taken before the first shock. These ECG segments were downsampled and then analyzed using three different metrics:

  • Fibrillation Power: This metric, adapted from previous work, quantifies the overall electrical energy of the VF signal.
  • Dominant Frequency: This is the highest peak in the spectral density of the ECG signal, indicating the most common oscillation frequency.
  • Mean Amplitude: This measures the average absolute deviation of the waveform from its mean, representing the signal's strength.

To quantitatively compare these predictors, the researchers used logistic regression to determine an optimal threshold for each metric. They then calculated sensitivity, specificity, positive predictive value, and negative predictive value for each threshold.

Datasets

The study used ECG data from 54 patients who experienced out-of-hospital cardiac arrest with ventricular fibrillation. This dataset was collected using an automatic external defibrillator (AED) at a sampling rate of 120 Hz. The paper does not introduce a new benchmark dataset but rather evaluates existing metrics on this patient data.

Key Insights and Ideas

  • Fibrillation Power as a Promising Predictor: The study demonstrates that "fibrillation power" is a valid predictor of defibrillation success, performing comparably to established metrics like dominant frequency and mean amplitude.
  • Complementary Information: The analysis revealed that fibrillation power, dominant frequency, and mean amplitude show little correlation with each other. This suggests they capture different aspects of the VF signal and can be used together in a multivariate approach for potentially enhanced predictive power.
  • Preference for Amplitude-Independent Metrics: The authors note that indicators independent of absolute amplitude are preferable due to the poor reproducibility of amplitude measurements. This makes fibrillation power and dominant frequency potentially more robust.

Key Results

  • Performance Comparison: All three predictors (fibrillation power, dominant frequency, and mean amplitude) showed similar performance in predicting defibrillation success, with dominant frequency having a slight edge.
  • Dominant frequency achieved a sensitivity of 79% and a negative predictive value of 76%.
  • Mean amplitude had a sensitivity of 71% and a negative predictive value of 70%.
  • Fibrillation power achieved a sensitivity of 64% and a negative predictive value of 67%.
  • Low Correlation: Figure 2 visually confirms the low correlation between the three indicators, supporting the idea that they provide complementary information.

Limitations and Future Work

  • Parameter Optimization: The parameters used to quantify fibrillation power were optimized based on retrospective porcine data. The authors suggest that further optimization of these parameters for human data could enhance predictive power.
  • Overall Unsatisfactory Performance: Despite the promising results for individual predictors, the overall performance in predicting defibrillation success is still considered unsatisfactory.
  • Need for Alternative Indicators: The development of alternative indicators, especially those that can analyze the ECG during ongoing CPR, should be pursued. The authors are confident that the VF ECG contains more exploitable information than is currently utilized.
  • Multivariate Analysis: The paper suggests that exploiting the complementary information from these indicators through multivariate discriminant analysis is a promising avenue for future research.

Summary updated

BibTeX
@inproceedings{hamprecht2001preliminary,
  author = {Fred A. Hamprecht and Daniel Jost and M Rüttimann and Franck Calamai and J Kowalski},
  title = {{Preliminary results on the prediction of countershock success with fibrillation power}},
  booktitle = {Resuscitation},
  year = {2001},
  doi = {10.1016/s0300-9572(01)00360-4},
}
Figure