High frequency oscillations in healthy brain functions

Abstract

Neural high frequency activity above 30 Hz has been linked to various brain functions, including sensory processing as well as higher-order functions like attention and memory. However, the origin, role, and function of this gamma band activity is still unclear and popular theories on high frequency activity are highly debated. This dissertation addresses two major aspects of gamma research: in the first part (Studies 1 and 2), it examines high frequency activity in the visual system, highlighting the interplay of retina and cortex. In the second part (Studies 3 and 4), it addresses the problem that gamma activity constitutes a rather weak signal by examining which recording conditions are ideal and whether the analysis of high frequency activity can benefit from state-of-the-art analysis methods.Study 1: Several studies imply that the processing of dark stimuli benefits from greater neural resources compared to the processing of light stimuli and is thus faster. Exactly which portions of the visual pathway could be involved in such differences is not yet resolved, and furthermore, related evidence from the human visual system is scant. This study examines the interplay of retina and cortex in the processing of darks and lights by simultaneously recording retinal and cortical responses with electroretinogram (ERG) and magnetoencephalography (MEG) to light offsets and onsets in ten participants. High frequency activity in response to light offset occurred faster than light onset in cerebral cortex, but not in the retina. Furthermore, the bandwidth of the onset and offset responses differed: while light onset elicited a broadband response, light offset was accompanied by narrowband gamma activity. The findings of this study suggest that retinal high frequency activity is transmitted to visual cortex, and that this transmission is presumably faster for light offset activity. These differences in propagation speed point to the importance of considering retinocortical interactions when interpreting cortical visual activity. Furthermore, this study contributes to the ongoing discussion about the origin and function of visual narrowband oscillations.Study 2: The retina clearly transfers massive amounts of information to visual cortex, but it is not conclusively resolved whether any information flows in the opposite direction in humans, from the cortex to the retina. This pilot study combines transcranial magnetic stimulation of visual cortex with the simultaneous recording of retinal activity to investigate whether the stimulation of cortical visual areas can affect the retina. In both subjects, retinal activity resembling flash-evoked activity was observed following transcranial stimulation of primary visual cortex, showing a slow potential as well as high frequency activity. Most of the suspected artifacts could be ruled out by sham stimulations and a phantom head investigation. The findings of this study are consistent with the existence of a corticofugal pathway and furthermore provide important indications for an improved design of the forthcoming full study.Study 3: The application of single-trial and decoding analyses can reveal meaningful brain activity that is obscured in the trial average. In the case of high frequency activity, however, the low signal-to-noise ratio complicates single-trial analyses. In this study, the applicability of a single-trial classification approach to decode stimulus modality from gamma activity was explored. The results show a successful classification of trials with auditory versus visual presentation of words across subjects. High frequency activity in both visual and auditory areas contributed to the classification model. Especially in visual cortex, this gamma activity had a broad bandwidth. The findings of this study show the feasibility of single-trial approaches to weak signals like high frequency activity and furthermore support the view that broadband and narrowband gamma activity may indeed have different roles and should be distinguished.Study 4: Source reconstruction with the beamforming technique is a widely used approach to localize brain activity and increase signal-to-noise ratio. Whether this approach profits from the growing number of channels in state-of-the-art recording systems, e.g. magnetoencephalographic systems, remains unclear. This study investigates how beamformer performance is impacted by sensor count in tandem with key properties of the input data, including signal strength. Counterintuitively, beamformer performance decreases with higher sensor count for strong input signals. With weak signals like high frequency activity, however, source reconstruction with beamformers improves with more sensors.Integrating these studies, the present thesis sheds light on the origin of high frequency activity in the visual system and highlights the importance of retinocortical interactions to the interpretation of visual cortical activity. It further provides new findings on the role of narrowband and broadband gamma activity and adds to the discussion how high frequency activity in the human brain may represent functional mechanisms. This work furthermore describes the impact of sensor count on beamformer performance, demonstrating that the reconstruction of weak signals like gamma activity profits from having more sensors. Finally, it demonstrates that decoding approaches, combined with beamforming, can successfully classify single-trial high frequency activity, with significant implications for cognitive applications and brain-computer interfaces.