The Application of Fractional Type of Blind Source Separation: Effect of Randomness in Interictal Epileptiform Discharges
Abstract number :
2.069
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2021
Submission ID :
1826346
Source :
www.aesnet.org
Presentation date :
12/5/2021 12:00:00 PM
Published date :
Nov 22, 2021, 06:53 AM
Authors :
Teppei Matsubara, MD, PhD - Athinoula A. Martinos Center for Biomedical Imaging; Sheraz Khan - Athinoula A. Martinos Center for Biomedical Imaging; Steven Stufflebeam - Athinoula A. Martinos Center for Biomedical Imaging; Matti Hämäläinen - Athinoula A. Martinos Center for Biomedical Imaging; Yoshinobu Goto - International University of Health and Welfare; Kuniharu Kishida - Gifu University
Rationale: In our previous paper, fractional type of blind source separation method (BSST/k) confirmed that interictal epileptiform discharges (IED) were extracted based on stochastic process (Clin Neurophysiol, 131; 425–36, 2020). BSST/k using time-delayed correlation takes account of the characteristic time structure, represented by the Fourier series of target signals (IEDs). The application of BSST/k has been established on the stationary process (IEEE Eng Med Biol Soc, 12; 299–329, 2013), where target signals periodically occur (e.g., evoked response). However, IEDs typically occur at random. The exact relationship between randomness and the spectrum of IEDs is unclear. Here, we used simulations to investigate the property of randomness which supports the theoretical background of the effectiveness of our BSST/k method in application to IEDs.
Methods: Magnetoencephalography data from 204 gradiometers (Elekta-neuromag, Helsinki, Finland) from one patient was investigated. A patient (Patient 5, left occipital lobe epilepsy) was randomly selected from our previous 7 patients with a single epileptogenic zone. IED occurred 59 counts/min. IED epochs (-20 – 180 ms) were chopped from the original raw data (240 sec) from one representative sensor (left occipital sensor). The remaining original data were concatenated to be used as the background. The randomly selected 100 IED epochs were embedded in the background according to the regularity of occurrence; continuous, constant, and random. Mathematically, the continuous pattern has characteristics of periodic functions. In constant and random patterns, the inter-IED interval (IED epoch had 200 ms duration) was variable from 20 ms (4.8 Hz) to 1000 ms (1.4 Hz), since 1000 ms inter-IED interval corresponds with real data (59 counts/min). Welch power spectral density estimate (PSD) was obtained. MNE-Python software was used for simulation as well as PSD calculation.
Results: In the continuous pattern, PSD showed clear line peaks that corresponded to the IED epoch (5 Hz). The constant pattern also showed clear line peaks that corresponded with inter-IED interval. In the random pattern, line peaks were still represented in fundamental and several harmonics either in 4.8 Hz and 4.1 Hz inter-IED interval, however, line peaks were not evident in 1.4 Hz inter-IED interval.
Conclusions: The randomness of occurrence of IED still keeps its spectrum similar to that of the constant pattern. In the longer inter-IED interval it should be considered that spectrum is hidden by the noise of the background, however, it still has its spectrum. This simulation data explains why non-stationary IEDs can be described by the way of perturbation approach within the weak stationary process. Furthermore, these results will be useful for the setting of T/k parameters, which may determine the waveforms of interest.
Funding: Please list any funding that was received in support of this abstract.: Nakatani Foundation for advancement of measuring technologies in biomedical engineering; Research Fellowships of Japan Society for the Promotion of Science for Young Scientists.
Neurophysiology