Removal of EOG Artifact in Electroencephalography with EEMD-ICA: A Semi-simulation Study on Identification of Artifactual Components

Purpose: The electroencephalography (EEG) signals recorded in clinical settings are usually corrupted by electrooculography (EOG) artifacts. EEMD-ICA is a commonly used method for removing EOG artifacts. This study aims at exploring the performance of different methods of identification of artifactual components under the framework of EEMD-ICA. Methods: This study is conducted in a semi-simulated way. A EEG dataset covering signal of SNR from -1 to 2 is generated based on the EEG and EOG segments from two public datasets. Characterized by the artifactual components identification method, EEMD-ICA \(_{kurt}\) , EEMD-ICA \(_{entropy}\) , EEMD-ICA \(_{autocor}\) and EEMD-ICA \(_{eogcor}\) are proposed and evaluated in terms of Normalized Mean Square Error (NMSE), Cross Correlation (CC) and Structural Similarity Index (SSIM) on this dataset. Results: EEMD-ICA \(_{autocor}\) outperforms other three approaches and demonstrates the strongest versatility. Besides successfully eliminating EOAs from EEG signals, it loses the least neuron activities. Conclusion: Although performance metrics improve as SNR increases, the loss of structure information also improves (SNR > 1). In practice, it is vital to estimate the SNR of data before applying these approaches because when SNR is high, these methods may have a counterproductive.