Comparison of synchrosqueezing transform to alternative methods for time-frequency analysis of TMS-evoked EEG oscillations

Abstract

BackgroundTranscranial magnetic stimulation combined with electroencephalography (TMS-EEG) has become a powerful tool to assess cortical properties, such as cortical oscillation. An accurate and robust time–frequency analysis tool with high resolution would facilitate the understanding of TMS-evoked oscillations. MethodsThe synchrosqueezing transform (SST), the Hilbert-Huang transform (HHT) and the Morlet wavelet transform (MWT) were used to analyze TMS-evoked oscillations. Firstly, we generated simulation data, and compared the performance of three methods to analyze the time–frequency characteristics of simulation data; then, we collected TMS-EEG data from normal people (NOR), patients with minimally conscious state (MCS) and vegetative state (VS). SST was used to analysis time–frequency characteristics of TMS-evoked oscillations in different states of consciousness. In addition, relative power (RP) and spectral entropy (SeEn) were calculated based on SST results. ResultsSimulation results showed that SST detected rhythm characteristics of instantaneous signal and background signal more completely. Results of NOR group showed that SST could more accurately detect TMS-evoked oscillations with high frequency resolution than HHT and MWT. The main frequency of TMS-evoked oscillations for DOC patients (MCS: 11.73 ± 1.94 Hz; VS: 2.06 ± 0.25 Hz) was different from that of NOR (22.79 ± 1.42 Hz) based on SST. RP and SeEn further verified the differences of the main frequency of TMS-evoked oscillations between NOR and DOC patients. ConclusionsThe results suggest that SST is a robust analytical method and outperforms HHT and MWT in studying TMS-evoked oscillations. The main frequency of TMS-evoked oscillations of DOC was lower than that of NOR.