Ulster University Logo

Use of Support Vector Machines in Predicting Success of Intracardiac Cardioversion by Electric Shocks in Patients with Atrial Fibrillation

Diaz, JD, Diaz, MA, Castro, NC, Glover, B, Manoharan, G and Escalona, OJ (2008) Use of Support Vector Machines in Predicting Success of Intracardiac Cardioversion by Electric Shocks in Patients with Atrial Fibrillation. In: IV LATIN AMERICAN CONGRESS ON BIOMEDICAL ENGINEERING 2007, BIOENGINEERING SOLUTIONS FOR LATIN AMERICA HEALTH, VOLS 1 AND 2. UNSPECIFIED. 5 pp. [Conference contribution]

[img] PDF - Published Version
Indefinitely restricted to Repository staff only.



The objective of this study, was to build support vector machines (SVM) for predicting success of electric shocks in the internal cardioversion (IC) of patients with atrial fibrillation (AF). Some investigations have found correlations between parameters and necessary energy for defibrillating AF, but no tool exist for predicting whether an electric shock will be successful or not in low energy IC. Thirty eight patients with AF, for elective DC cardioversion at the Royal Victoria Hospital in Belfast, were included in our study. Two catheters were positioned in the right atrial appendage (RAA) and the coronary sinus (CS), to deliver a biphasic shock waveform, synchronized with the R wave of the electrocardiogram (ECG) signal. A voltage step-up protocol (50-300 V) was used for patient cardioversion. The ECG was analyzed for an average time interval of 52,8 +/- 10.1 s (corresponding to segments before each shock). Residual atrial fibrillatory signal (RAFS) was estimated by means of bandpass filtering and ventricular activity (QRST) cancellation. QRST complexes were cancelled using a recursive least squared (RLS) adaptive filter. The atria[ fibrillatory frequency (AFF) and the instantaneous frequency (IF) series were extracted from the RAFS. AFF was calculated from whole segments and from the 10 seconds of the RAFS previous shocks. The mean, standard deviation and approximate entropy of the IF time series were computed. RR intervals of the ECG segments were also analyzed. A total of 26 patients were successfully cardioverted, employing 167 shocks (141 non successful). SVMs were built for classifying success on shocks for energy up to 2, 3 and 6 Joules, employing different combinations of the computed parameters. A maximal exactitude of 93.42% (sensitivity=92.31% and specificity=93.65%) was obtained classifying shocks below 2 Joules, 95.51% (sensitivity=92.86% and specificity=96%) for shocks up to 3 Joules, and 92.91% (sensitivity=78.95% and specificity=95.37%) for shocks <= 6 Joules.

Item Type:Conference contribution (Paper)
Keywords:Support vector machine; cardioversion; atrial fibrillatory frequency; instantaneous frequency; QRS cancellation; defibrillation
Faculties and Schools:Faculty of Computing & Engineering
Faculty of Computing & Engineering > School of Engineering
Research Institutes and Groups:Engineering Research Institute
Engineering Research Institute > Nanotechnology & Integrated BioEngineering Centre (NIBEC)
ID Code:6078
Deposited By: Professor Omar Escalona
Deposited On:14 Jan 2010 09:12
Last Modified:09 Dec 2015 10:04

Repository Staff Only: item control page