Search bioRxivSearch

Biology subjects

Leech, R.

Publications and source records attributed to Leech, R..

6 recordsLinked to original sources

Tracking and optimizing human performance using deep reinforcement learning in closed-loop behavioral- and neurofeedback: a proof of concept

Reinforcement learning (RL) is a general-purpose powerful machine learning framework within which we can model various deterministic, non-deterministic and complex environments. We applied RL to the problem of tracking and improving human sustained attention during a simple sustained attention to response task (SART) in a proof of concept study with two subjects, using state-of-the-art deep neural network-based RL in the form of Deep Q Networks (DQNs). While others have used RL in EEG settings previously, none have applied it in a neurofeedback (NFB) setting, which seems a natural problem within Brain Computer Interfaces (BCIs) to tackle using end-to-end RL in the form of DQNs, due to both the problems non-stationarity and the ability of RL to learn in a continuous setting. Furthermore, while many have explored phasic alerting previously, learning optimal alerting in a personalized way in real time is a less explored field, which we believe RL to be an most suitable solution for. First, we used empirically-derived simulated data of EEG and reaction times and subsequent parameter/algorithmic exploration within this simulated model to pick parameters for the DQN that are more likely to be optimal for the experimental setup and to explore the behavior of DQNs in this task setting. We then applied the method on two subjects and show that we get different but plausible results for both subjects, suggesting something about the behavior of DQNs in this setting. For this experimental part, we used parameters suggested to us by the simulation results. This RL-based behavioral- and neuro-feedback BCI method we have developed here is input feature agnostic and allows for complex continuous actions to be learned in other more complex closed-loop behavioral or neuro-feedback approaches.

neuroscience

Bayesian Optimisation for Neuroimaging Pre-processing in Brain Age Prediction

Neuroimaging-based age predictions using machine learning have been shown to relate to cognitive performance, health outcomes and progression of neurodegenerative disease. However, even leading age-prediction algorithms contain measurement error, motivating efforts to improve experimental pipelines. T1-weighted MRI is commonly used for age prediction, and the pre-processing of these scans involves normalisation to a common template and resampling to a common voxel size, followed by spatial smoothing. Resampling parameters are often selected arbitrarily. Here, we sought to improve brain-age prediction accuracy by optimising resampling parameters using Bayesian optimisation.\n\nUsing data on N=2001 healthy individuals (aged 16-90 years) we trained support vector machines to i) distinguish between young (<50 years) and old (>50 years) brains and ii) predict chronological age, with accuracy assessed using cross-validation. We also evaluated model generalisability to the Cam-CAN dataset (N=648, aged 18-88 years). Bayesian optimisation was used to identify optimal voxel size and smoothing kernel size for each task. This procedure adaptively samples the parameter space to evaluate accuracy across a range of possible parameters, using independent sub-samples to iteratively assess different parameter combinations to arrive at optimal values.\n\nWhen distinguishing between young and old brains a classification accuracy of 96.25% was achieved, with voxel size = 11.5mm3 and smoothing kernel = 2.3mm. For predicting chronological age, a mean absolute error (MAE) of 5.08 years was achieved, with voxel size = 3.73mm3 and smoothing kernel = 3.68mm. This was compared to performance using default values of 1.5mm3 and 4mm respectively, which gave a MAE = 5.48 years, a 7.3% improvement. When assessing generalisability, best performance was achieved when applying the entire Bayesian optimisation framework to the new dataset, out-performing the parameters optimised for the initial training dataset.\n\nOur study demonstrates the proof-of-principle that neuroimaging models for brain age prediction can be improved by using Bayesian optimisation to select more appropriate pre-processing parameters. Our results suggest that different parameters are selected and performance improves when optimisation is conducted in specific contexts. This motivates use of optimisation techniques at many different points during the experimental process, which may result in improved statistical sensitivity and reduce opportunities for experimenter-led bias.

neuroscience

Brain state and polarity dependent modulation of brain networks by transcranial direct current stimulation

Transcranial direct current stimulation (TDCS) has been widely used to improve cognitive function. However, current deficiencies in mechanistic understanding hinders wider applicability. To clarify its physiological effects, we acquired fMRI whilst simultaneously acquiring TDCS to the right inferior frontal gyrus (rIFG) of healthy human participants, a region involved in coordinating activity within brain networks. TDCS caused widespread modulation of network activity depending on brain state ( rest or choice reaction time task) and polarity (anodal or cathodal). During task, TDCS increased salience network activation and default mode network deactivation, but had the opposite effect during rest. Furthermore, there was an interaction between brain state and TDCS polarity, with cathodal effects more pronounced during task performance and anodal effects more pronounced during rest. Overall, we show that rIFG TDCS produces brain state and polarity dependent effects within large-scale cognitive networks, in a manner that goes beyond predictions from the current literature.

neuroscience

Effects of LSD on music-evoked brain activity

Music is a highly dynamic stimulus, and consists of distinct acoustic features, such as pitch, rhythm and timbre. Neuroimaging studies highlight a hierarchy of brain networks involved in music perception. Psychedelic drugs such as lysergic acid diethylamide (LSD) temporary disintegrate the normal hierarchy of brain functioning, and produce profound subjective effects, including enhanced music-evoked emotion. The primary objective of this study was to investigate the acute effects of LSD on music-evoked brain-activity under naturalistic music listening conditions. 16 healthy participants were enrolled in magnetic resonance imaging (fMRI) while listening to a 7-minute music piece under eyes-closed conditions on two separate visits (LSD (75 mcg) and placebo). Dynamic time courses for acoustic features were extracted from the music excerpts, and were entered into subject-level fMRI analyses as regressors of interest. Differences between conditions were assessed at group level subsequently, and were related to changes in music-evoked emotions via correlation analyses. Psycho-physiological interactions (PPIs) were carried out to further interrogate underlying music-specific changes in functional connectivity under LSD. Results showed pronounced cortical and subcortical changes in music-evoked brain activity under LSD. Most notable changes in brain activity and connectivity were associated with the component timbral complexity, representing the complexity of the musics spectral distribution, and these occurred in brain networks previously identified for music-perception and music-evoked emotion, and showed an association with enhanced music-evoked feelings of wonder under LSD. The findings shed light on how the brain processes music under LSD, and provide a neurobiological basis for the usefulness of music in psychedelic therapy.

neuroscience

Assessing tACS-induced phosphene perception using closed-loop Bayesian optimization

Transcranial alternating current stimulation (tACS) can evoke illusory flash-like visual percepts known as phosphenes. The perception of phosphenes represents a major experimental challenge when studying tACS-induced effects on cognitive performance. Besides growing concerns that retinal phosphenes themselves could potentially have neuromodulatory effects, the perception of phosphenes may also modify the alertness of participants. Past research has shown that stimulation intensity, frequency and electrode montage affect phosphene perception. However, to date, the effect of an additional tACS parameter on phosphene perception has been completely overlooked: the relative phase difference between stimulation electrodes. This is a crucial and timely topic given the confounding nature of phosphene perception and the increasing number of studies reporting changes in cognitive function following tACS phase manipulations. However, studying phosphene perception for different frequencies and phases simultaneously is not tractable using standard approaches, as the physiologically plausible range of parameters results in a combinatorial explosion of experimental conditions, yielding impracticable experiment durations. To overcome this limitation, here we applied a Bayesian optimization approach to efficiently sample an exhaustive tACS parameter space. Moreover, unlike conventional methodology, which involves subjects judging the perceived phosphene intensity on a rating scale, our study leveraged the strength of human perception by having the optimization driven based on a subjects relative judgement. Applying Bayesian optimization for two different montages, we found that phosphene perception was affected by differences in the relative phase between cortical electrodes. The results were replicated in a second study involving new participants and validated using computational modelling. In summary, our results have important implications for the experimental design and conclusions drawn from future tACS studies investigating the effects of phase on cognition.

neuroscience

Dissociating frontoparietal brain networks with neuroadaptive Bayesian optimization

Understanding the unique contributions of frontoparietal networks (FPN) in cognition is challenging because different FPNs spatially overlap and are co-activated for diverse tasks. In order to characterize these networks involves studying how they activate across many different cognitive tasks, which previously has only been possible with meta-analyses. Here, building upon meta-analyses as a starting point, we use neuroadaptive Bayesian optimization, an approach combining real-time analysis of functional neuroimaging data with machine-learning, to discover cognitive tasks that dissociate ventral and dorsal FPN activity from a large pool of tasks. We identify and subsequently refine two cognitive tasks (Deductive Reasoning and Tower of London) that are optimal for dissociating the FPNs. The identified cognitive tasks are not those predicted by meta-analysis, highlighting a different mapping between cognitive tasks and FPNs than expected. The optimization approach converged on a similar neural dissociation independently for the two different tasks, suggesting a possible common underlying functional mechanism and the need for neurally-derived cognitive taxonomies.

neuroscience