Abstract 273: A Novel Method for Detection of QRS Complex With an Adaptive Points Insertion
Introduction: An important step in many ECG analysis techniques is extraction of the QRS complex. Existing QRS complex detection algorithms utilize methods such as neural networks, genetic algorithms, wavelet transform, filter banks, and heuristic methods. Many of these algorithms provide low accuracies for irregular ECGs and cannot be implemented in real-time making them unsuitable for applications such as resuscitations. Detection errors can be divided into two types: false detection and missing detection. We developed and tested a novel signal processing technique utilizing Hilbert transform in order to improve QRS complex detection. The technique decreases the missing detection, ensuring that the false detection is as small as possible.
Methods: The proposed method is designed to search for the R wave in each cycle. The Hilbert transform is applied to the original ECG thus forming a new signal in which not only the amplitude of the difference between the R wave and other waves is greatly enhanced, but also in which inverted R waves are identified. A novel threshold method that produces the minimum errors between two different maximal points is then applied to obtain the most likely R waves. Finally, an adaptive process is used to exclude possible false R waves. In the correcting stage, amplitude and period controllers are applied. The period controller helps in deciding whether to insert a missing R wave between two successive detected R waves, while the amplitude controller checks whether there are possible R waves satisfying the insertion condition. In the end, the adaptive process excludes possible false R waves. The MIT-BIH arrhythmia data bank was used to test the performance.
Results: Forty-eight case datasets were utilized in testing. The technique resulted in an accuracy, sensitivity, and specificity of 99.8%, of 99.9% of 99.89%, respectively. Moreover, the point insertion process produced correct placement in 99.10% of cases.
Conclusions: The designed algorithm detects the QRS complex with a very high accuracy. Not only can the algorithm overcome the disadvantages of existing QRS complex detection algorithms to a large extent, but also its simplicity makes it suitable for real-time implementation for applications such as resuscitation.
- © 2010 by American Heart Association, Inc.