The Quality Assessment Of Dynamic ECG Based On The Fusion Of Multi-index

Cardiovascular diseases rank the most common reason of deaths resulting from non-communicable diseases.Since there are no obvious early symptoms,diseases progress worse before suitable examination and treatment are conducted,which can easily lead to a sudden death,therefore,effective means of prevention is essential.Electrocardiogram(ECG)is a comprehensive reflection of the heart's electrical activity on the surface,and it is also helpful to analyze cardiac-related pathologies.Dynamic electrocardiogram,which can record ECG signal of human behavior for 24 hours or even longer.can capture sudden diseases and reduce morbidity rate.However,ECG acquired dynamically will be affected by environmental and behavioral factors too much to guarantee its quality.On one hand,the affected signal quality is not beneficial to the diagnosis and treatment of related diseases;on the other hand,a large number of invalid information increases the workload and waste of resources.Therefore,before processing and analysis,the quality assessment of dynamic ECG is of vital importance to scientific research and practical value.Having thoroughly conducted the research of ECG quality assessment technology,this thesis proposed one quality assessment method based on multi-index fusion,mainly using dynamic ECG signals actually collected from the lab.The method is composed of three parts,index extraction,fusion and quality assessment.Indices can be divided into outer layer index and inner layer index:the former one is electrode fall-off detection index proposed by electrode fall-off experiment;the latter one is extracted from the analysis of time domain and frequency domain,including baseline wander level index based on wavelet transform,amplitude change level index based on the difference between top and bottom cenvelopes transform,and the frequency component index based on power spectrum estimation.As to fusion,this thesis conducted additive synthesis based on entropy weight to the fusion of inner layer indices,and multiplicative synthesis to the fusion of inner and outer layer indices.As to quality assessment,the outer layer index classifies signals as acceptable and unacceptable while the inner indices classifies available signals into good and bad using the threshold acquired from the interval estimation method.For the results,made a comparison between it and the experiential threshold method on the basis of manual evaluation:the average accuracy and the average false positive rate of interval estimation method are 93.29%and 94.49%,as for the experiential threshold method,they are 85.96%and 88.14%.Unlike instabilities originated from threshold adjustments,this method is robust to diverse data.