Cardiac magnetic resonance image-based classification of the risk of arrhythmias in post-myocardial infarction patients
作者：
Highlights：
• Cardiac magnetic resonance image-based features were used to distinguish post-myocardial infarction patients into high and low arrhythmic risk groups.
• Seventeen features describing the size, location and texture of the scarred myocardium were used in different classifiers.
• In Experiment 1, a systematic testing of features and their combinations was done.
• SMOTE, wrapper based feature selection, and nested cross-validation were used in Experiment 2.
• Experiments 1 and 2 gave an accuracy of 94.4% (AUC = 0.965) and 92.6% (AUC = 0.921), respectively.
摘要
Highlights•Cardiac magnetic resonance image-based features were used to distinguish post-myocardial infarction patients into high and low arrhythmic risk groups.•Seventeen features describing the size, location and texture of the scarred myocardium were used in different classifiers.•In Experiment 1, a systematic testing of features and their combinations was done.•SMOTE, wrapper based feature selection, and nested cross-validation were used in Experiment 2.•Experiments 1 and 2 gave an accuracy of 94.4% (AUC = 0.965) and 92.6% (AUC = 0.921), respectively.
论文关键词：Local binary pattern,Sobel filter,k-Nearest neighbor classifier,Support vector machine classifier,Cardiac magnetic resonance image,High and low arrhythmic risk
论文评审过程：Received 30 September 2013, Revised 8 June 2015, Accepted 25 June 2015, Available online 4 July 2015, Version of Record 2 September 2015.
论文官网地址：https://doi.org/10.1016/j.artmed.2015.06.001
