Abstract:
The interaction of different neural networks is essential for cognition and for healthy brain function, and the cross-frequency interactions between different neuronal oscillations have been hypothesized to coordinate such interactions. However, little is understood about the distribution and the nature of the cross-frequency relationships in the human brain.
A prominently researched mode of cross-frequency interaction has been phase-amplitude coupling (PAC). Here, the phase of a low frequency oscillation is associated with the amplitude of a high frequency oscillation. As a methodological complication, however, measures for PAC do not only reflect the presence of a phase-to-amplitude interaction between two distinct oscillations, but they also reflect the presence of a neuronal single oscillation with a non-sinusoidal waveform. In the case of a non-sinusoidal waveform, the measured cross-frequency relationship reflects the wave shape of the oscillation. The underlying mechanisms and functions of two distinct oscillations that are phase-amplitude coupled to one-another differ from those of a a single oscillation with a particular waveform. Therefore, a clear dissociation between the two cases is essential to gain a meaningful understanding of the cross-frequency relationships in the human brain.
In the first study of this thesis, we systematically mapped PAC across the human cortex, and in a wide range of frequency pairs using magnetoencephalography (MEG) and source reconstruction. We distinguished neuronal PAC from non-neuronal PAC related to muscle activity and eye-movements, and we showed that phase-amplitude, phase-phase (PPC), and amplitude-amplitude (AAC) cross-frequency coupling measures are all sensitive to signals with higher harmonics. We used these measures in conjunction to dissociate non-harmonic and harmonic PAC. We found no evidence of non-harmonic PAC in the resting human brain. Instead, we observed widespread PAC that was driven by harmonic signals, predominantly in the alpha frequency range. That is, we observed widespread alpha oscillations with non-sinusoidal wave shape.
The results of the first study raised the question whether alpha oscillations in different brain areas may have different wave shapes. That is, if the wave shape of oscillations observed in human resting state MEG might be functionally relevant. To address this question in the second study, we first, determined spatial peaks of theta/alpha wave-shape stability using MEG, source-reconstruction and bicoherence. Then we assessed the wave shapes at these regions of interest (ROIs) with a novel method. With this method, wave shapes were analyzed in the frequency domain, by exploiting the characteristic cross-frequency patterns of signals with higher harmonics. We tested for wave-shape differences and distinguished six statistically different alpha wave shapes: three corresponding to the well-established functionally distinct sensorimotor-, occipital- and temporal alpha rhythms, and three additional parietal alpha waveforms.
These studies, to our best knowledge for the first time, systematically characterized the distribution and the nature of cross-frequency signals in the resting human brain. We showed that non-sinusoidal wave shapes were a prevalent phenomenon in the human cortex that dominated all observable cross-frequency patterns. Furthermore, we demonstrated that the characteristic cross-frequency patterns of non-sinusoidal wave shapes can be used to differentiate what are likely functionally distinct rhythms. Periodic wave shapes can be reconstructed in detail from their characteristic cross-frequency patterns, and the wave shape of oscillations might reflect rich information about underlying circuit physiology.
cross-frequency coupling