Digital filters for low-latency quantification of brain rhythms in real time

Feedback control of collective dynamics in an oscillator population with time-dependent connectivity

We present a numerical study of pulsatile feedback-based control of synchrony level in a highly-interconnected oscillatory network. We focus on a nontrivial case when the system is close to the synchronization transition point and exhibits collective rhythm with strong amplitude modulation. We pay special attention to technical but essential steps like causal real-time extraction of the signal of interest from a noisy measurement and estimation of instantaneous phase and amplitude. The feedback loop's parameters are tuned automatically to suppress synchrony. Though the study is motivated by neuroscience, the results are relevant to controlling oscillatory activity in ensembles of various natures and, thus, to the rapidly developing field of network physiology.

Real-time low latency estimation of brain rhythms with deep neural networks

Objective.Neurofeedback and brain-computer interfacing technology open the exciting opportunity for establishing interactive closed-loop real-time communication with the human brain. This requires interpreting brain's rhythmic activity and generating timely feedback to the brain. Lower delay between neuronal events and the appropriate feedback increases the efficacy of such interaction. Novel more efficient approaches capable of tracking brain rhythm's phase and envelope are needed for scenarios that entail instantaneous interaction with the brain circuits.Approach.Isolating narrow-band signals incurs fundamental delays. To some extent they can be compensated using forecasting models. Given the high quality of modern time series forecasting neural networks we explored their utility for low-latency extraction of brain rhythm parameters. We tested five neural networks with conceptually distinct architectures in forecasting synthetic EEG rhythms. The strongest architecture was then trained to simultaneously filter and forecast EEG data. We compared it against the state-of-the-art techniques using synthetic and real data from 25 subjects.Main results.The temporal convolutional network (TCN) remained the strongest forecasting model that achieved in the majority of testing scenarios>90% rhythm's envelope correlation with<10 ms effective delay and<20∘circular standard deviation of phase estimates. It also remained stable enough to noise level perturbations. Trained to filter and predict the TCN outperformed the cFIR, the Kalman filter based state-space estimation technique and remained on par with the larger Conv-TasNet architecture.Significance.Here we have for the first time demonstrated the utility of the neural network approach for low-latency narrow-band filtering of brain activity signals. Our proposed approach coupled with efficient implementation enhances the effectiveness of brain-state dependent paradigms across various applications. Moreover, our framework for forecasting EEG signals holds promise for investigating the predictability of brain activity, providing valuable insights into the fundamental questions surrounding the functional organization and hierarchical information processing properties of the brain.

An adaptive acoustoelectric signal decoding algorithm based on Fourier fitting for brain function imaging

Acousticelectric brain imaging (ABI), which is based on the acoustoelectric (AE) effect, is a potential brain function imaging method for mapping brain electrical activity with high temporal and spatial resolution. To further enhance the quality of the decoded signal and the resolution of the ABI, the decoding accuracy of the AE signal is essential. An adaptive decoding algorithm based on Fourier fitting (aDAF) is suggested to increase the AE signal decoding precision. The envelope of the AE signal is first split into a number of harmonics by Fourier fitting in the suggested aDAF. The least square method is then utilized to adaptively select the greatest harmonic component. Several phantom experiments are implemented to assess the performance of the aDAF, including 1-source with various frequencies, multiple-source with various frequencies and amplitudes, and multiple-source with various distributions. Imaging resolution and decoded signal quality are quantitatively evaluated. According to the results of the decoding experiments, the decoded signal amplitude accuracy has risen by 11.39% when compared to the decoding algorithm with envelope (DAE). The correlation coefficient between the source signal and the decoded timing signal of aDAF is, on average, 34.76% better than it was for DAE. Finally, the results of the imaging experiment show that aDAF has superior imaging quality than DAE, with signal-to noise ratio (SNR) improved by 23.32% and spatial resolution increased by 50%. According to the experiments, the proposed aDAF increased AE signal decoding accuracy, which is vital for future research and applications related to ABI.

Real-time phase and amplitude estimation of neurophysiological signals exploiting a non-resonant oscillator

A recent advancement in the field of neuromodulation is to adapt stimulation parameters according to pre-specified biomarkers tracked in real-time. These markers comprise short and transient signal features, such as bursts of elevated band power. To capture these features, instantaneous measures of phase and/or amplitude are employed, which inform stimulation adjustment with high temporal specificity. For adaptive neuromodulation it is therefore necessary to precisely estimate a signal's phase and amplitude with minimum delay and in a causal way, i.e. without depending on future parts of the signal. Here we demonstrate a method that utilizes oscillation theory to estimate phase and amplitude in real-time and compare it to a recently proposed causal modification of the Hilbert transform. By simulating real-time processing of human LFP data, we show that our approach almost perfectly tracks offline phase and amplitude with minimum delay and is computationally highly efficient.

Real-time estimation of phase and amplitude with application to neural data

Computation of the instantaneous phase and amplitude via the Hilbert Transform is a powerful tool of data analysis. This approach finds many applications in various science and engineering branches but is not proper for causal estimation because it requires knowledge of the signal's past and future. However, several problems require real-time estimation of phase and amplitude; an illustrative example is phase-locked or amplitude-dependent stimulation in neuroscience. In this paper, we discuss and compare three causal algorithms that do not rely on the Hilbert Transform but exploit well-known physical phenomena, the synchronization and the resonance. After testing the algorithms on a synthetic data set, we illustrate their performance computing phase and amplitude for the accelerometer tremor measurements and a Parkinsonian patient's beta-band brain activity.

Short-delay neurofeedback facilitates training of the parietal alpha rhythm

OBJECTIVE

Feedback latency was shown to be a critical parameter in a range of applications that imply learning. The therapeutic effects of neurofeedback (NFB) remain controversial. We hypothesized that often encountered unreliable results of NFB intervention could be associated with large feedback latency values that are often uncontrolled and may preclude the efficient learning.

APPROACH

We engaged our subjects into a parietal alpha power unpregulating paradigm faciliated by visual neurofeedback based on the invidually extracted envelope of the alpha-rhythm at P4 electrode. NFB was displayed either as soon as EEG envelope was processed, or with an extra 250 or 500-ms delay. The feedback training consisted of 15 two-minute long blocks interleaved with 15s pauses. We have also recorded two minute long baselines immediately before and after the training.

MAIN RESULTS

The time course of NFB-induced changes in the alpha rhythm power clearly depended on NFB latency, as shown with the adaptive Neyman test. NFB had a strong effect on the alpha-spindle incidence rate, but not on their duration or amplitude. The sustained changes in alpha activity measured after the completion of NFB training were negatively correlated to latency, with the maximum change for the shortest tested latency and no change for the longest.

SIGNIFICANCE

Here we for the first time show that visual NFB of parietal electroencephalographic (EEG) alpha-activity is efficient only when delivered to human subjects at short latency, which guarantees that NFB arrives when an alpha spindle is still ongoing. Such a considerable effect of NFB latency on the alpha-activity temporal structure could explain some of the previous inconsistent results, where latency was neither controlled nor documented. Clinical practitioners and manufacturers of NFB equipment should add latency to their specifications while enabling latency monitoring and supporting short-latency operations.