Room: AAPM ePoster Library
In this paper a novel electrocardiogram (ECG) signals classification approach is proposed based on Local Binary Patterns (LBPs). Although LBP is one of the new and very simple methods used in texture classification but in this paper the method is applied in ECG classification.
The training phase consists of applying LBP operator to all different ECG signals individually to extract the reference feature vectors for each class of heart disease. Consequently, each signal is windowed in classification phase and LBP operator is applied to each window and compared with all reference feature vectors. To increase the accuracy of the method two further morphological features are investigated, which are Variance and Mean.
The proposed method is applied to eleven different classes of signals selected from MIT-BIH dataset and the average accuracy of 99.76 is obtained. While no signal pre-processing step is required and considering the simplicity and accuracy of the proposed signal classification algorithm, it can be considered as a robust one which is suitable for online ECG classification.
Considering the low computational complexity of the proposed algorithm, it can be used as a suitable algorithm for online ECG classification too. According to simplicity and satisfactorily accurate rate of classification, it is evident that wider range of arrhythmia can be considered and classified by the proposed algorithm. On the other hand, interesting features of LBP can trigger researchers to apply the method in other medical signal and image processing applications.
Electrocardiograms, Signal Processing, Classifier Design
IM/TH- Image Analysis Skills (broad expertise across imaging modalities): Computer-aided decision support systems (detection, diagnosis, risk prediction, staging, treatment response assessment/monitoring, prognosis prediction)