Tunable Q-wavelet based ECG data compression with validation using cardiac arrhythmia patterns

The acquisition of longer electrocardiogram (ECG) signals is an essential step of the recent medical screening and diagnostic procedures. Considering the need of an efficient ECG data management system, this paper presents a novel ECG data compression technique based on the tunable Q-wavelet transform which provides adjustable parameters to achieve good compression performance. The tunable Q-wavelet transform compacts the maximum energy of the signal to fewer transform coefficients. Dead-zone quantization is used to discard the small valued transform coefficients. Further, transform coefficients are rounded-off to nearest integer values and encoded using run-length coding. ECG records of the MIT-BIH arrhythmia database are used to evaluate the performance of the proposed compression technique. The average compression performance obtained in terms of CR, PRD, PRD1, QS, QS(1) and SNR are 20.61, 4.43, 6.37, 5.88, 3.46, and 55.93 dB respectively. The proposed technique offers better compression performance than many existing techniques. The proposed technique is also validated using cardiac arrhythmia patterns classification. In the validation phase, tunable Q-wavelet transform based features of 14,878 original cardiac patterns are used to train the support vector machine classifier while testing is performed using features of 26,219 original and reconstructed cardiac patterns separately. It examines the effect of the proposed compression technique on cardiac arrhythmia classification. In the validation part, the overall classification accuracy, sensitivity, and specificity obtained for the proposed compression technique are 98.35%, 95.77%, and 99.19% respectively. It denotes that the proposed technique compresses ECG signal efficiently with preserving diagnostic information very well.

Automated summary

The paper presents a novel ECG data compression technique based on the tunable Q-wavelet transform, achieving good compression performance by compacting signal energy and discarding small valued transform coefficients. Experimental results show that the proposed technique performs well in cardiac arrhythmia classification.