Abstract:
Electrocardiogram (ECG) datasets are among the most challenging records that have been widely studied for early automatic prediction of cardiac anomalies. In order to ach...Show MoreMetadata
Abstract:
Electrocardiogram (ECG) datasets are among the most challenging records that have been widely studied for early automatic prediction of cardiac anomalies. In order to achieve high performance automatic prediction, existing works make use of complex and time consuming techniques and/or show high rates of false positives. In this paper, we introduce a new method to analyze an ECG dataset and perform an efficient prediction of 7 ST-segment and T-wave anomalies related to Myocardial Infarction (MI) or Ischemia. Our method combines both Decision Trees Boosting and Random Under Sampling (RUS) techniques to respectively improve the prediction performance and solve the class imbalance problem. This method, named RUSBoost, has been validated using data of 7 leads, collected from a real ECG dataset [1], and the obtained results show a higher balance between true and false positives for all the 7 leads. Obtained average sensitivity and specificity are respectively 86% and 94.85%, which outperform the existing results of other related works.
Published in: 2016 IEEE 18th International Conference on e-Health Networking, Applications and Services (Healthcom)
Date of Conference: 14-16 September 2016
Date Added to IEEE Xplore: 21 November 2016
ISBN Information: