Abstract 8782: Spectral Techniques Distinguish Clinical Ventricular Fibrillation from Artifact in Implantable Cardioverter-Defibrillators
Introduction: Several studies have shown that 11–24% of patients with implantable cardioverter-defibrillators (ICD) receive inappropriate shocks. Etiologies include lead artifact (AR) from electromagnetic interference and lead malfunction. Prior work has shown that spectral techniques can distinguish AR from induced ventricular fibrillation (VF) during defibrillation threshold testing, but limited work has been done to see if these techniques distinguish clinical VF from clinical AR.
Methods: A retrospective analysis was performed on 32 clinical VF and 35 clinical AR events from 27 patients that were detected by an implanted ICD. AR events included artifact from both lead malfunction and electromagnetic interference. All events were retrieved from the storage diskette and analyzed off-line using Matlab software. A power spectral density analysis was performed on a 2-second segment of the far-field data for both VF and AR. An Energy Ratio (ER), the sum of the energy of the first 3 harmonics to the energy of the entire spectrum, was calculated for both VF and AR. A receiver operating characteristic curve (ROC) was generated to assess the ability of ER to distinguish VF from AR. Data are expressed as mean ± standard deviation.
Results: The Energy Ratio of VF was significantly larger than the Energy Ratio of AR (0.8545 ± 0.1119 vs. 0.3441 ± 0.1283, p < 0.0001). The ROC Area under the Curve was 0.996 (CI 0.939–1.000). Using a cutoff of 0.581, the Energy Ratio had a sensitivity of 96.9% and a specificity of 97.1% to distinguish VF from AR. The dominant frequency for AR was significantly larger for AR than VF (16.8 ± 26.4 vs. 4.8 ± 1.3 Hz, p = 0.01).
Conclusions: Signal obtained during VF has unique spectral characteristics that may be used to distinguish VF from EMI or lead malfunction.
- © 2010 by American Heart Association, Inc.