Article

Prediction of countershock success using single features from multiple ventricular fibrillation frequency bands and feature combinations using neural networks

Department of Anaesthesiology and Critical Care Medicine, Innsbruck Medical University, Innsbruck, Austria.
Resuscitation (Impact Factor: 3.96). 06/2007; 73(2):253-63. DOI: 10.1016/j.resuscitation.2006.10.002
Source: PubMed

ABSTRACT Targeted defibrillation therapy is needed to optimise survival chances of ventricular fibrillation (VF) patients, but at present VF analysis strategies to optimise defibrillation timing have insufficient predictive power. From 197 patients with in-hospital and out-of-hospital cardiac arrest, 770 electrocardiogram (ECG) recordings of countershock attempts were analysed. Preshock VF ECG features in the time and frequency domain were tested retrospectively for outcome prediction. Using band pass filters, the ECG spectrum was split into various frequency bands of 2-26 Hz bandwidth in the range of 0-26 Hz. Neural networks were used for single feature combinations to optimise prediction of countershock success. Areas under curves (AUC) of receiver operating characteristics (ROC) were used to estimate prediction power of single and combined features. The highest ROC AUC of 0.863 was reached by the median slope in the interval 10-22 Hz resulting in a sensitivity of 95% and a specificity of 50%. The best specificity of 55% at the 95% sensitivity level was reached by power spectrum analysis (PSA) in the 6-26 Hz interval. Neural networks combining single predictive features were unable to increase outcome prediction. Using frequency band segmentation of human VF ECG, several single predictive features with high ROC AUC>0.840 were identified. Combining these single predictive features using neural networks did not further improve outcome prediction in human VF data. This may indicate that various simple VF features, such as median slope already reach the maximum prediction power extractable from VF ECG.

0 Followers
 · 
94 Views
  • Source
    • "Earlier analysis has presented promising results in discriminating between PEA and perfusing rhythm[2]. The following parameters were adapted from prior work and used in this study: average RR interval, number of detected QRS complexes, average QRS width, average QRS height, average ECG power, average ECG amplitude, average ECG amplitude exceeding 80% of maximum amplitude (MA80), angle, slope (average sample difference), signal length (SL) of the minimum phase correspondent (MPC), max MPC, form factor and the coefficient of a 6 th order polynomial fitting with additional fitting error[2] [3] [4] [5]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Possible clinical states of a cardiac arrest patient are ventricular fibrillation/tachycardia (VF/VT), asystole (ASY) or pulseless electrical activity (PEA), and the treatment goals are return of spontaneous circulation (ROSC) and neurologic ally intact survival. Waveform analysis has been used in VF to predict treatment outcomes and we hypothesised that similar analysis in PEA could predict transformation to ROSC. We analysed 120 and 83 PEA segments prior to transitions to ROSC and ASY, respectively, to investigate the ability often electrocardiograph (ECG) features to predict transitions to ROSC or ASY using neural networks. The feature combination that yielded the best discrimination had a meanplusmnSD area under the receiver operating characteristics curve of 0.88plusmn0.02. The results suggest that the ECG contains information regarding the dynamics of PEA which can be used to study effects of therapies in cardiac arrest patients.
    Computers in Cardiology, 2006; 10/2006
  • Source
  • [Show abstract] [Hide abstract]
    ABSTRACT: Several different ventricular fibrillation (VF) analysis features based on ECG have been reported for shock outcome prediction. In this study we investigated the influence of the time from VF onset to shock delivery (VF duration) and the rhythm before onset of VF, on the probability of return of spontaneous circulation (ROSC). We also analysed how these factors relate to the VF analysis feature median slope. ECG recordings from 221 cardiac arrest patients from previously published prospective studies on the quality of CPR were used. VF duration and prior rhythm were determined when VF occurred during the episode. Median slope before each shock was calculated. The median VF duration was shorter in shocks producing ROSC, 24 seconds (s) versus 70s (P<0.001). VF duration shorter than 30s resulted in 27% ROSC versus 10% for those longer than 30s (OR=3.5 [95% CI: 2.2-5.4]). The prior rhythm influenced the probability of ROSC, with perfusing rhythm being superior, followed by PEA, asystole, and "poor" PEA (broad complexes and/or irregular/very slow rate), respectively. The probability of ROSC corresponded well with the average median slope value for each group, but the correlation between median slope and VF duration was very poor (r2=0.05). Based on our findings, detection of VF during ongoing chest compressions might be valuable because VF of short duration was associated with ROSC. Further, the rhythm before VF affects shock outcome with a perfusing rhythm giving the best prospect. The median slope can be used for shock outcome prediction, but not for determining VF duration. A combination could be beneficial and warrants further studies.
    Resuscitation 10/2007; 75(1):60-7. DOI:10.1016/j.resuscitation.2007.02.014 · 3.96 Impact Factor
Show more