2006 · 14 citations · 4 references
EngineeringBiomedical EngineeringCardiologyBiomedical Signal ProcessingElectrophysiologyMachine LearningCardiac Signal ProcessingClassificationDecision TreesElectrocardiogram AnalysisElectrocardiographySupport Vector MachinesClassification ModelsMyocardial Infarction DetectionNaive BayesNaive Bayes ClassificationRandom Forest ClassificationSupervised ClassificationSupervised Classification AlgorithmsSupport Vector Machine Classification
In this study we have investigated the classification of old myocardial infarction through the analysis of 192 lead Body Surface Potential Maps (BSPM). Following an analysis of the most prominent features based on a signal to noise ratio ranking criterion the top 6 features were selected. These features were subsequently used as inputs to a series of supervised classification models in the form of Naïve Bayes (NB), Support Vector Machine (SVM) and Random Forest (RF)-based classifiers. Following 10-fold cross validation it was found that the best performance for each classifier was 81.9 % for NB, 82.8 % for SVM and 84.5 % for RF. The results have indicated the ability of the approach to successfully classify the recordings based on a non standard subset of recording sites from the BSPM. 1.
4