Abstract 17882: A Method for Real-time CPR Detection From the Thoracic Impedance Signal
Introduction: Accurate detection of chest compressions is a means to improve CPR through real-time feedback or post-event review. Accelerometer technology is the gold standard to detect compressions but is not universally available and requires additional equipment.
Hypothesis: An algorithm derived from the transthoracic impedance signal could achieve high sensitivity and specificity for compression detection and accurately measure the CPR fraction.
Methods: Defibrillator recordings from cardiac arrest cases were used to develop (N=20) and validate (N=62) the compression detection algorithm. Accelerometer data were used to define the presence or absence of compressions at 1-second intervals throughout entire cases, up to 10 minutes per case. The impedance-based algorithm cross-correlates the impedance signal with a series of wavelets and then integrates sequential outputs with a hidden Markov model. Using accelerometer classification as the gold standard, the sensitivity and specificity of the impedance-based algorithm for compression detection were estimated by a multilevel model to account for clustering of classifications within cases. The difference in CPR fractions was calculated for each case by subtracting the accelerometer-derived CPR fraction from the impedance-derived CPR fraction, and their median and 95% CI were estimated by the Wilcoxon signed-rank test.
Results: The validation set included a median 535 seconds per case. The impedance-derived algorithm had mean sensitivity 0.95 (95% CI 0.93, 0.96) and mean specificity 0.91 (95% CI 0.90, 0.93). Median CPR fraction was 0.74 (IQR 0.65, 0.79) based on the accelerometer assessment. The impedance-derived CPR fraction was 0.71 (95% CI 0.64, 0.77), for a difference of -0.01 (95% CI -0.03, 0.00).
Conclusion: A method that identifies CPR from the impedance signal alone had high mean sensitivity and high mean specificity but slightly underestimated the CPR fraction.
Author Disclosures: H. Kwok: None. J. Coult: None. L. Sherman: None. J. Blackwood: None. P. Kudenchuk: None. T. Rea: None.
- © 2015 by American Heart Association, Inc.