Support vector machine for arrhythmia discrimination with wavelet transform-based feature selection

Support Vector Machines (SVMs), have meant a great advance in solving classification or pattern recognition problems. The present contribution is devoted to applying SVM to malignant arrhythmias discrimination. The Wavelet Transform was applied to single-lead episodes of different rhythms belonging to various patients. More than 50 characteristic parameters were extracted in order to define each rhythm. The number of normalized parameters were reduced by means of backward algorithms developed by the authors. SVM was then applied to the reduced normalized parameter set. SVM surpassed other classification schemes, including advanced statistical decision methods. Good-accuracy classifications are achieved with just a few support vectors, with the consequent benefit in computational cost. In conclusion, these positive results evidence the potential of SVM techniques in malignant arrhythmias discrimination