Spike isolation from background signal in neonatal EEG data using an integrated independent component analysis method
Conference
Asl, ME, Rodriguez, JD, Cabrerizo, M et al. (2024). Spike isolation from background signal in neonatal EEG data using an integrated independent component analysis method
. 2014 36TH ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY (EMBC), 10.1109/EMBC53108.2024.10782611
Asl, ME, Rodriguez, JD, Cabrerizo, M et al. (2024). Spike isolation from background signal in neonatal EEG data using an integrated independent component analysis method
. 2014 36TH ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY (EMBC), 10.1109/EMBC53108.2024.10782611
Spike detection in epileptic neonates is a challenging task since the recorded electroencephalography (EEG) data are often fraught with artifacts and noise. This study aims to enhance the clarity of epileptic spikes by separating them from background activity using an integrated method based on the independent component analysis (ICA). We analyzed spikes of 12 epileptic neonates as marked in their EEG scalp recordings by a clinical expert. The proposed method makes use of the ICA method to isolate the source of the spikes and then apply a power frequency analysis and template matching to validate the performance of the ICA. Isolating a spike is achieved by choosing the component that should correspond to its defining characteristics, followed by signal reconstruction using that component. To evaluate the accuracy of our spike isolation method, we first check if the power spectrum of the separated spikes aligns with the typical power spectral density observed in neonates. Subsequently, we measured the degree of similarity between the extracted spike and a predefined spike template, comparing it against the original spike segment. With this integrated method, the results show the successful extraction of 29 out of the 37 marked spikes (i.e., 79 percent), which signifies that ICA can serve as a promising approach in the initial isolation process of spikes in EEG records of neonates. This could lead to further investigation into those subtle features or changes missed on those EEG records of the marked spikes that were not separated. Determining such features and subtle changes, if indeed inherent to spikes, could lead to the development of enhanced spike detection methods in neonates. It should be noted that in 5 out of the 37 epochs, we could not identify any independent component as a spike source, and 3 out of 32 remaining cases showed unsuccessful separation in validation, possibly due to the source not being statistically independent or being Gaussian in nature. In such cases, the expert clinician(s) could review or reconsider marking such spikes.
