Introduction There is accumulating evidence that brain function is strongly dependent on dynamic interactions between neural assemblies. Several methods, based on synchronization measures, are used to quantify functional connectivity between neural networks. However, despite the dynamic nature of neuronal interactions, functional connectivity is typically expressed as a mean value during the period of observation. In the present work, we characterize and quantify the spatiotemporal dynamics of these interactions using common instantaneous frequencies between different recording positions. We explore this approach in two conditions: wakefulness and generalized anesthesia. Methods We used 19-channel EEGs from patients who underwent a carotid endarterectomy during general anesthesia with propofol. Pre-operative 20 min EEGs were also available. All recordings were obtained as part of our routine care. All EEGs were recorded with Ag/AgCl electrodes with a sampling frequency of 250 Hz. EEGs were decomposed in the time-frequency plane by a short time Fourier transform. Subsequently, we used the ridge transform to subtract instantaneous frequencies from the local maxima in the time-frequency plane. The resulting ridge plots allow analysis of common instantaneous frequencies between different oscillators. We now define a link rate (LR) and a link duration (LD) for quantifying the frequency and duration of interactions, respectively. We illustrate this approach for different frequency bands (delta, theta, alpha and beta), using the EEG recordings during the eyes closed wake state and general anesthesia. Results EEGs from 21 patients were used. Link rates in the theta, alpha and beta band were significantly higher during anesthesia than during wakefulness (3.65–5.34 Hz during anesthesia, 3.51–4.55 Hz during wakefulness, p ≪ 0.05), while the link rates in the delta band were significantly lower during anesthesia (2.49 Hz during anesthesia, 2.96 Hz during wakefulness, p ≪ 0.05). The link durations in the delta and beta band were significantly shorter during anesthesia (158–124 ms during anesthesia, 168–129 ms during wakefulness, p ≪ 0.05), while those in the theta and alpha band were significantly higher (151–169 ms during anesthesia, 149–160 ms during wakefulness, p ≪ 0.05). Conclusion We present a method to characterize dynamic interactions between neural networks in the brain. A next step is to include phase information to limit volume conduction effects. Despite this limitation, we further show that link durations in most frequency bands decrease during anesthesia, supporting the notion that in this condition ‘neural integration’ is reduced.
Read full abstract