We applied Ensemble Empirical Mode Decomposition (EEMD) to extract low-frequency, theta and supra-theta signals from raw LFPs (Figure S1 ) using the PSTA package (https://pennmem.github.io/ptsa_new/html/index.html ). The EEMD consists of breaking down a time varying, non-stationary signal into its elementary signals referred to as the Intrinsic Mode Functions (IMFs) by iteratively applying the empirical mode decomposition algorithm with added white noise to prevent mode mixing (Wu and Huang, 2009 ). We extracted the theta signal of each raw LFP by combining the IMFs with mean instantaneous frequencies between 5 and 12 Hz. Low-frequency and supra-theta signals were defined as the sum of IMFs with mean frequencies below 5 Hz and above 12 Hz, respectively (Figure S1 ). Note that we used EEMD to obtain the theta waveform and avoid harmonic artifacts related to cycle asymmetries. Besides being an unsupervised filter (i.e., free of predefined frequency bands), one of the main advantage of the EEMD is that it deals well with asymmetrical (non-linear) and non-stationary signals, thus diminishing filtering artifacts (such as harmonics and side band-related distortions) caused by convolution filters for cross-frequency coupling analysis (Aru et al., 2015 (link), Belluscio et al., 2012 (link), Yeh et al., 2016 (link)). Therefore, apart from having the theta signal automatically extracted from the raw LFP, the EEMD also provides supra-theta components that are virtually free from harmonic artifacts (Wu and Huang, 2009 ).
Free full text: Click here
