Quality Assessment Of ECG Data By Channel

The conventional Holter usually records ECG data for 24 hours or even longer,during which the user’s breathing,body movement and the deterioration of electromagnetic environment may reduce the quality of ECG recording.Before diagnosis,we should find out the low-quality data with no medical reference value,such as serious noise interference,and remove them in advance.This requires an assessment of the quality of the recorded data.This method can be used not only to evaluate the data quality of single channel ECG equipment,but also to evaluate the data quality of multi-channel ECG equipment.The quality of maternal ECG components in the abdominal wall of pregnant women recorded in fetal ECG monitoring may be high,but the quality of the remaining fetal ECG components after removing maternal ECG components may not be high or available,and there is also a problem of evaluating the quality of separated fetal ECG signals.In this paper,the methods of data quality evaluation of conventional ECG and fetal ECG by channel are studied.The main work is as follows:(1)For conventional ECG signals,a single channel quality assessment method is proposed.A new classification method is defined.According to the waveform characteristics of each channel,it can be classified into four categories: electrode shedding,serious noise interference,partial noise interference and high-quality signal.The waveform and frequency domain characteristics of conventional ECG signal are analyzed in detail,and six quality indexes are proposed.The normalization and dimension reduction of index matrix are combined with one-class support vector machine to evaluate the channel signal quality.The experimental results show that the classification accuracy of electrode shedding is 93.22%,that of serious noise interference is 90%,that of partial noise interference is 89.22%,and that of high quality signal is 97.19%.(2)For fetal ECG signal,a single channel quality assessment method is proposed.Firstly,the abdominal wall ECG signals are pre classified.Using the method in(1),the data classified as partial noise interference and high quality signal are retained,and the fetal ECG is extracted from these data.Each channel is labeled as acceptable or unacceptable according to the characteristics of fetal ECG waveform.The linear and nonlinear characteristics of FECG are analyzed,and eight quality indexes are proposed.At the same time,one-class support vector machine is used to evaluate the quality of fetal channel signal.The experimental results show that the final classification accuracy is 92.14%.When the self-test database is used as the test set,the classification accuracy is 90.34%.(3)To study the quality of fetal ECG synthesis signal.Principal component analysis(PCA),independent component analysis(ICA),repetitive component analysis(RCA)and periodic component analysis(πCA)are used to enhance each fetal ECG data,and one FECG signal is synthesized from multiple signals for quality analysis.The experimental results show that the accuracy of PCA,ICA,RCA and πCA is 89.43%,90.86%,93.67% and 92.59%.Compared with the classification results before enhancement,it has better accuracy.