Scalp EEG Quantitative Quality Assessment Method From A Big Date Perspective And Its Application

In the EEG applications,the real EEG signal may be masked by various types of artifacts.Although there are many artifact removal methods,high-quality raw EEG data is still irreplaceable.Therefore,the quality assessment method for the raw EEG data is particularly significant and has benefits as following:1)timely feedback on data quality during data acquisition;2)exclusion of poor-quality data;3)development of personalized preprocessing strategies based on quality assessment results.However,the existing studys on EEG quality assessment still has some limitations(e.g.,scattered quantitative indices,insufficient generalization and low robustness).Based on the characteristics of raw EEG data,a few robust,quantitative and objective quality indices were developed,and a standardized quality assessment method was established by statistically analyzing large-scale window indices.The main work of this thesis contains:Firstly,a new Quantitative Signal Quality Assessment method(QSQA)aiming at raw EEG data from a big data perspective was proposed;and then a simulation dataset with different noise types and noise levels was established to test the performance of QSQA,and an eye-closed resting-state healthy subject dataset was used to test and optimize parameters,explore the optimal window indice thresholds and investigate the relationship between parameters and QSQA results.For the simulated data,it was found that the quality indices showed a linear trend(R~2>0.99)as the noise level increased,indicating QSQA was sensitive for artifacts.For the resting-state dataset,the suggested thresholds corresponded to the 95%reference interval of the window indice distributions,and the data quality indices remained stable when the parameters exceeded the suggested parameters,indicating stable performance of QSQA under the suggested parameters.Then,the proposed method was compared with the PREP tool for performance difference,and four resting-state EEG datasets were used to further explore the application of QSQA.The results showed that compared with PREP tool,QSQA could more finely and accurately assess the EEG data quality.For resting-state EEG datasets,it was found that frequency noise was more common than other artifact types(under the quality level"D",the OFN of three datasets are 34.72%,51.97%and 44.92%respectively),which suggested that special attention should be paid to the frequency noise interference in the EEG data acquisition and preprocessing.In summary,QSQA was sensitivie to noise,and a series of stable and reliable quantitative indices of EEG data quality were proposed,which had a significant application prospect in large-scale EEG research.In addition,QSQA has been integrated into the We Brain cloud platform(https://webrain.uestc.edu.cn/).