• Trygve Eftestøl, Heidrun Losert, Jo Kramer-Johansen, Lars Wik, Fritz Sterz, and Petter Andreas Steen.
• University of Stavanger, Faculty of Science and Technology, Department of Electrical and Computer Engineering, N-4036 Stavanger, Norway. trygve.eftestol@uis.no
• Resuscitation. 2005 Oct 1; 67 (1): 55-61.

AbstractWe evaluated the ability of a previously derived outcome predictor to discriminate between ECG segments corresponding to return of spontaneous circulation (ROSC) or not in validation data from 136 patients with cardiac arrest. The new data used for validation were totally independent from the predictor derivation data used in the original study. Features corresponding to those used in the development of the original outcome predictor, centroid frequency, peak power frequency, spectral flatness and energy, were computed following which a second decorrelated feature set was generated. The outcome predictor was applied to the new data with good correspondence in performance (testing) to what was expected (training) with receiver operator characteristics (ROC) areas of 0.80 and 0.79, respectively. Outcome predictor performance was reproducible. As in the present study, future testing should be performed on totally independent data not included in the design of the outcome predictor to get a reliable impression of expected performance.

26 keywords...

hide…