A support vector machine for predicting defibrillation outcomes from waveform metrics.

Andrew Howe, OJ Escalona, Rebecca Di Maio, Bertrand Massot, Nick A Cromie, Karen M Darragh, Jennifer Adgey, DJ McEneaney

Research output: Contribution to journalArticle

8 Citations (Scopus)

Abstract

Background: Algorithms to predict shock success based on VF waveform metrics could significantly enhance resuscitation by optimising the timing of defibrillation. Objective: To investigate robust methods of predicting defibrillation success in VF cardiac arrest patients, by using a support vector machine (SVM) optimisation approach. Methods: Frequency-domain (AMSA, dominant frequency and median frequency) and time-domain (slope and RMS amplitude) VF waveform metrics were calculated in a 4.1s window prior to defibrillation. Conventional prediction test validity of each waveform parameter was conducted and used AUC > 0.6 as the criterion for inclusion as a corroborative attribute processed by the SVM classification model. The latter used a Gaussian radial-basis-function (RBF) kernel and the error penalty factor C was fixed to 1. A two-fold cross-validation resampling technique was employed. Results: A total of 41 patients had 115 defibrillation instances. AMSA, slope and RMS waveform metrics performed test validation with AUC > 0.6 for predicting termination of VF and return-to-organised rhythm. Predictive accuracy of the optimised SVM design for termination of VF was 81.9% (±1.24 SD); positive and negative predictivity were respectively 84.3% (±1.98 SD) and 77.4% (±1.24 SD); sensitivity and specificity were 87.6% (±2.69 SD) and 71.6% (±9.38 SD) respectively. Conclusions: AMSA, slope and RMS were the best VF waveform frequency–time parameters predictors of termination of VF according to test validity assessment. This a priori can be used for a simplified SVM optimised design that combines the predictive attributes of these VF waveform metrics for improved prediction accuracy and generalisation performance without requiring the definition of any threshold value on waveform metrics.
LanguageEnglish
Pages343-349
JournalResuscitation
Volume85
Issue number3
DOIs
Publication statusPublished - 28 Nov 2013

Fingerprint

Amsacrine
Area Under Curve
Heart Arrest
Resuscitation
Shock
Sensitivity and Specificity
Support Vector Machine

Keywords

  • Cardiac arrest
  • defibrillation
  • VF waveform metrics
  • group classification
  • resuscitation therapy
  • outcome prediction
  • support vector machine
  • SVM
  • supervised machine learning
  • CPR

Cite this

Howe, Andrew ; Escalona, OJ ; Di Maio, Rebecca ; Massot, Bertrand ; Cromie, Nick A ; Darragh, Karen M ; Adgey, Jennifer ; McEneaney, DJ. / A support vector machine for predicting defibrillation outcomes from waveform metrics. In: Resuscitation. 2013 ; Vol. 85, No. 3. pp. 343-349.
@article{348e911ea7c54731a3d6e94f744295ac,
title = "A support vector machine for predicting defibrillation outcomes from waveform metrics.",
abstract = "Background: Algorithms to predict shock success based on VF waveform metrics could significantly enhance resuscitation by optimising the timing of defibrillation. Objective: To investigate robust methods of predicting defibrillation success in VF cardiac arrest patients, by using a support vector machine (SVM) optimisation approach. Methods: Frequency-domain (AMSA, dominant frequency and median frequency) and time-domain (slope and RMS amplitude) VF waveform metrics were calculated in a 4.1s window prior to defibrillation. Conventional prediction test validity of each waveform parameter was conducted and used AUC > 0.6 as the criterion for inclusion as a corroborative attribute processed by the SVM classification model. The latter used a Gaussian radial-basis-function (RBF) kernel and the error penalty factor C was fixed to 1. A two-fold cross-validation resampling technique was employed. Results: A total of 41 patients had 115 defibrillation instances. AMSA, slope and RMS waveform metrics performed test validation with AUC > 0.6 for predicting termination of VF and return-to-organised rhythm. Predictive accuracy of the optimised SVM design for termination of VF was 81.9{\%} (±1.24 SD); positive and negative predictivity were respectively 84.3{\%} (±1.98 SD) and 77.4{\%} (±1.24 SD); sensitivity and specificity were 87.6{\%} (±2.69 SD) and 71.6{\%} (±9.38 SD) respectively. Conclusions: AMSA, slope and RMS were the best VF waveform frequency–time parameters predictors of termination of VF according to test validity assessment. This a priori can be used for a simplified SVM optimised design that combines the predictive attributes of these VF waveform metrics for improved prediction accuracy and generalisation performance without requiring the definition of any threshold value on waveform metrics.",
keywords = "Cardiac arrest, defibrillation, VF waveform metrics, group classification, resuscitation therapy, outcome prediction, support vector machine, SVM, supervised machine learning, CPR",
author = "Andrew Howe and OJ Escalona and {Di Maio}, Rebecca and Bertrand Massot and Cromie, {Nick A} and Darragh, {Karen M} and Jennifer Adgey and DJ McEneaney",
year = "2013",
month = "11",
day = "28",
doi = "10.1016/j.resuscitation.2013.11.021",
language = "English",
volume = "85",
pages = "343--349",
journal = "Resuscitation",
issn = "0300-9572",
publisher = "Elsevier",
number = "3",

}

A support vector machine for predicting defibrillation outcomes from waveform metrics. / Howe, Andrew; Escalona, OJ; Di Maio, Rebecca; Massot, Bertrand; Cromie, Nick A; Darragh, Karen M; Adgey, Jennifer; McEneaney, DJ.

In: Resuscitation, Vol. 85, No. 3, 28.11.2013, p. 343-349.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A support vector machine for predicting defibrillation outcomes from waveform metrics.

AU - Howe, Andrew

AU - Escalona, OJ

AU - Di Maio, Rebecca

AU - Massot, Bertrand

AU - Cromie, Nick A

AU - Darragh, Karen M

AU - Adgey, Jennifer

AU - McEneaney, DJ

PY - 2013/11/28

Y1 - 2013/11/28

N2 - Background: Algorithms to predict shock success based on VF waveform metrics could significantly enhance resuscitation by optimising the timing of defibrillation. Objective: To investigate robust methods of predicting defibrillation success in VF cardiac arrest patients, by using a support vector machine (SVM) optimisation approach. Methods: Frequency-domain (AMSA, dominant frequency and median frequency) and time-domain (slope and RMS amplitude) VF waveform metrics were calculated in a 4.1s window prior to defibrillation. Conventional prediction test validity of each waveform parameter was conducted and used AUC > 0.6 as the criterion for inclusion as a corroborative attribute processed by the SVM classification model. The latter used a Gaussian radial-basis-function (RBF) kernel and the error penalty factor C was fixed to 1. A two-fold cross-validation resampling technique was employed. Results: A total of 41 patients had 115 defibrillation instances. AMSA, slope and RMS waveform metrics performed test validation with AUC > 0.6 for predicting termination of VF and return-to-organised rhythm. Predictive accuracy of the optimised SVM design for termination of VF was 81.9% (±1.24 SD); positive and negative predictivity were respectively 84.3% (±1.98 SD) and 77.4% (±1.24 SD); sensitivity and specificity were 87.6% (±2.69 SD) and 71.6% (±9.38 SD) respectively. Conclusions: AMSA, slope and RMS were the best VF waveform frequency–time parameters predictors of termination of VF according to test validity assessment. This a priori can be used for a simplified SVM optimised design that combines the predictive attributes of these VF waveform metrics for improved prediction accuracy and generalisation performance without requiring the definition of any threshold value on waveform metrics.

AB - Background: Algorithms to predict shock success based on VF waveform metrics could significantly enhance resuscitation by optimising the timing of defibrillation. Objective: To investigate robust methods of predicting defibrillation success in VF cardiac arrest patients, by using a support vector machine (SVM) optimisation approach. Methods: Frequency-domain (AMSA, dominant frequency and median frequency) and time-domain (slope and RMS amplitude) VF waveform metrics were calculated in a 4.1s window prior to defibrillation. Conventional prediction test validity of each waveform parameter was conducted and used AUC > 0.6 as the criterion for inclusion as a corroborative attribute processed by the SVM classification model. The latter used a Gaussian radial-basis-function (RBF) kernel and the error penalty factor C was fixed to 1. A two-fold cross-validation resampling technique was employed. Results: A total of 41 patients had 115 defibrillation instances. AMSA, slope and RMS waveform metrics performed test validation with AUC > 0.6 for predicting termination of VF and return-to-organised rhythm. Predictive accuracy of the optimised SVM design for termination of VF was 81.9% (±1.24 SD); positive and negative predictivity were respectively 84.3% (±1.98 SD) and 77.4% (±1.24 SD); sensitivity and specificity were 87.6% (±2.69 SD) and 71.6% (±9.38 SD) respectively. Conclusions: AMSA, slope and RMS were the best VF waveform frequency–time parameters predictors of termination of VF according to test validity assessment. This a priori can be used for a simplified SVM optimised design that combines the predictive attributes of these VF waveform metrics for improved prediction accuracy and generalisation performance without requiring the definition of any threshold value on waveform metrics.

KW - Cardiac arrest

KW - defibrillation

KW - VF waveform metrics

KW - group classification

KW - resuscitation therapy

KW - outcome prediction

KW - support vector machine

KW - SVM

KW - supervised machine learning

KW - CPR

UR - http://www.elsevier.com/locate/resuscitation

U2 - 10.1016/j.resuscitation.2013.11.021

DO - 10.1016/j.resuscitation.2013.11.021

M3 - Article

VL - 85

SP - 343

EP - 349

JO - Resuscitation

T2 - Resuscitation

JF - Resuscitation

SN - 0300-9572

IS - 3

ER -