Support vector machine and hidden Markov model based decoding of finger movements using electrocorticography

Authors:

• Tobias Wissel

• Tim Pfeiffer

• Robert Frysch

• Robert T. Knight

• Edward F. Chang

• Hermann Hinrichs

• Jochem W. Rieger

• Georg Rose

Date: 2013

DOI: 10.1088/1741-2560/10/5/056020

View PDF

Abstract:

Objective. Support vector machines (SVM) have developed into a gold standard for accurate classification in brain–computer interfaces (BCI). The choice of the most appropriate classifier for a particular application depends on several characteristics in addition to decoding accuracy. Here we investigate the implementation of hidden Markov models (HMM) for online BCIs and discuss strategies to improve their performance. Approach. We compare the SVM, serving as a reference, and HMMs for classifying discrete finger movements obtained from electrocorticograms of four subjects performing a finger tapping experiment. The classifier decisions are based on a subset of low-frequency time domain and high gamma oscillation features. Main results. We show that decoding optimization between the two approaches is due to the way features are extracted and selected and less dependent on the classifier. An additional gain in HMM performance of up to 6% was obtained by introducing model constraints. Comparable accuracies of up to 90% were achieved with both SVM and HMM with the high gamma cortical response providing the most important decoding information for both techniques. Significance. We discuss technical HMM characteristics and adaptations in the context of the presented data as well as for general BCI applications. Our findings suggest that HMMs and their characteristics are promising for efficient online BCIs.

Network dynamics predict improvement in working memory performance following Donezepil administration in health young controls

Authors:

• Amit Reches

• Ilan Laufer

• Guy Cukierman

• Kevin McEvoy

• M. Ettinger

Date: 2013

DOI: 10.1016/j.neuroimage.2013.11.020

PubMed: 24269569

View PDF

Abstract:

Attentional selection in the context of goal-directed behavior involves top-down modulation to enhance the contrast between relevant and irrelevant stimuli via enhancement and suppression of sensory cortical activity. Acetylcholine (ACh) is believed to be involved mechanistically in such attention processes. The objective of the current study was to examine the effects of donepezil, a cholinesterase inhibitor that increases synaptic levels of ACh, on the relationship between performance and network dynamics during a visual working memory (WM) task involving relevant and irrelevant stimuli. Electroencephalogram (EEG) activity was recorded in 14 healthy young adults while they performed a selective face/scene working memory task. Each participant received either placebo or donepezil (5mg, orally) on two different visits in a double-blinded study. To investigate the effects of donepezil on brain network dynamics we utilized a novel EEG-based Brain Network Activation (BNA) analysis method that isolates location-time-frequency interrelations among event-related potential (ERP) peaks and extracts condition-specific networks. The activation level of the network modulated by donepezil, reflected in terms of the degree of its dynamical organization, was positively correlated with WM performance. Further analyses revealed that the frontal-posterior theta-alpha sub-network comprised the critical regions whose activation level correlated with beneficial effects on cognitive performance. These results indicate that condition-specific EEG network analysis could potentially serve to predict beneficial effects of therapeutic treatment in working memory.

Role of the lateral prefrontal cortex in speech monitoring

Authors:

• Stephanie Ries

• Kira Xie

• Kathleen Y. Haaland

• Nina F. Dronkers

• Robert T. Knight

Date: 2013

DOI: 10.3389/fnhum.2013.00703

View PDF

Abstract:

The role of lateral prefrontal cortex (LPFC) in speech monitoring has not been delineated. Recent work suggests that medial frontal cortex (MFC) is involved in overt speech monitoring initiated before auditory feedback. This mechanism is reflected in an event- related potential (ERP), the error negativity (Ne), peaking within 100 ms after vocal-onset. Critically, in healthy individuals the Ne is sensitive to the accuracy of the response; it is larger for error than correct trials. By contrast, patients with LPFC damage are impaired in non-verbal monitoring tasks showing no amplitude difference between the Ne measured in correct vs. error trials. Interactions between the LPFC and the MFC are assumed to play a necessary role for normal action monitoring. We investigated whether the LPFC was involved in speech monitoring to the same extent as in non-linguistic actions by comparing performance and EEG activity in patients with LPFC damage and in aged-matched controls performing linguistic (Picture Naming) and non-linguistic (Simon) tasks. Controls did not produce enough errors to allow the comparison of the Ne or other ERP in error vs. correct trials. PFC patients had worse performance than controls in both tasks, but their Ne was larger for error than correct trials only in Naming. This task-dependent pattern can be explained by LPFC-dependent working-memory requirements present in non-linguistic tasks used to study action monitoring but absent in picture naming. This suggests that LPFC may not be necessary for speech monitoring as assessed by simple picture naming. In addition, bilateral temporal cortex activity starting before and peaking around vocal-onset was observed in LPFC and control groups in both tasks but was larger for error than correct trials only in Naming, suggesting the temporal cortex is associated with on-line monitoring of speech specifically when access to lexical representations is necessary.

Phase-amplitude cross-frequency coupling in the human nucleus acumbens tracks action monitoring during cognitive control

Authors:

• S. Durschmid

• Tino Zaehle

• Klaus Kopitzki

• Jürgen Voges

• Friedhelm Schmitt

• Hans-Jochen Heinze

• Robert T. Knight

• Hermann Hinrichs

Date: 2013

DOI: 10.3389/fnhum.2013.00635

PubMed: 24586885

View PDF

Abstract:

The Nucleus Accumbens (NAcc) is an important structure for the transfer of information between cortical and subcortical structures, especially the prefrontal cortex and the hippocampus. However, the mechanism that allows the NAcc to achieve this integration is not well understood. Phase-amplitude cross-frequency coupling (PAC) of oscillations in different frequency bands has been proposed as an effective mechanism to form functional networks to optimize transfer and integration of information. Here we assess PAC between theta and high gamma oscillations as a potential mechanism that facilitates motor adaptation. To address this issue we recorded intracranial field potentials directly from the bilateral human NAcc in three patients while they performed a motor learning task that varied in the level of cognitive control needed to perform the task. As in rodents, PAC was observable in the human NAcc, transiently occurring contralateral to a movement following the motor response. Importantly, PAC correlated with the level of cognitive control needed to monitor the action performed. This functional relation indicates that the NAcc is engaged in action monitoring and supports the evaluation of motor programs during adaptive behavior by means of PAC.

Error-monitoring and post-error compensations: dissociation between perceptual failures and motor errors with and without awareness

Authors:

• Robert T. Knight

• Ana Navarro-Cebrian

• Andrew S. Kayser

Date: 2013

DOI: 10.1523/JNEUROSCI.0447-13.2013

PubMed: 23884943

View PDF

Abstract:

Whether humans adjust their behavior in response to unaware errors remains a controversial issue relevant to insight in neuropsychiatric conditions. Initial error awareness studies found that the error-related negativity (ERN), an event-related potential (ERP) originating in the medial prefrontal cortex after errors, activated equally for aware and unaware errors, suggesting a candidate preconscious mechanism. However, recent studies demonstrate that the ERN decreases after unaware errors. We hypothesized that the ERN is dependent upon awareness, and predicted that previous discrepancies might be due to unaware errors not being differentiated from perceptually uncertain, low-confidence responses that might increase the ERN amplitude. Here we addressed this hypothesis by distinguishing between aware errors, unaware errors, and uncertain responses, and using stimuli (faces) associated with well established sensory ERPs to evaluate the degree of stimulus processing for each trial type. We found that while aware and unaware errors were related to failures at the time of response, uncertain responses were due to failures at the time of stimulus processing indexed by lower amplitude sensory ERPs. Moreover, uncertain responses showed similar ERN activity as aware errors, in comparison with decreased activity for unaware errors. Finally, compared with aware errors, uncertain responses and unaware errors showed reduced neural compensations, such as alpha suppression. Together these findings suggest that the ERN is activated by aware motor errors as well as sensory failures, and that both awareness and certainty are necessary for neural adaptations after errors.

Preparatory attention after lesions to the lateral or orbital prefrontal cortex: an event-related potential study

Authors:

• Ingrid Funderud

• Marianne Lovstad

• Magnus Lindgren

• Tor Endestad

• Paulina Due-Tønnesse

• Torstein Meling

• Robert T. Knight

• Anne-Kristin Solbakk

Date: 2013

PubMed: 23831520

View PDF

Abstract:

The prefrontal cortex (PFC) plays a central role in preparatory and anticipatory attentional processes. To investigate whether subregions of the PFC play differential roles in these processes we investigated the effect of focal lesions to either lateral prefrontal (lateral PFC; n=11) or orbitofrontal cortex (OFC; n=13) on the contingent negative variation (CNV), an electrophysiological index of preparatory brain processes. The CNV was studied using a Go/NoGo delayed response task where an auditory S1 signaled whether or not an upcoming visual S2 was a Go or a NoGo stimulus. Neither early (500–1000 ms) nor late (3200–3700 ms) phase Go trial CNV amplitude was reduced for any of the patient groups in comparison to controls. However, the lateral PFC group showed enhanced Go trial early CNV and reduced late CNV Go/NoGo differentiation. These data suggests that normal orienting and evaluation as reflected by the CNV is intact after OFC lesions. The enhanced early CNV after lateral PFC damage may be due to failure in inhibition and the reduced late CNV difference wave confirms a deficit in preparatory attention after damage to this frontal subregion.

Independence of valence and reward in emotional word processing: Electrophysiological evidence

Authors:

• Laura Kaltwasser

• Stephanie Ries

• Werner Sommer

• Robert T. Knight

• Roel M. Willems

Date: 2013

DOI: 10.3389/fpsyg.2013.00168

View PDF

Abstract:

Both emotion and reward are primary modulators of cognition: Emotional word content enhances word processing, and reward expectancy similarly amplifies cognitive processing from the perceptual up to the executive control level. Here, we investigate how these primary regulators of cognition interact. We studied how the anticipation of gain or loss modulates the neural time course (event-related potentials, ERPs) related to processing of emotional words. Participants performed a semantic categorization task on emotional and neutral words, which were preceded by a cue indicating that performance could lead to monetary gain or loss. Emotion-related and reward-related effects occurred in different time windows, did not interact statistically, and showed different topographies. This speaks for an independence of reward expectancy and the processing of emotional word content. Therefore, privileged processing given to emotionally valenced words seems immune to short-term modulation of reward. Models of language comprehension should be able to incorporate effects of reward and emotion on language processing, and the current study argues for an architecture in which reward and emotion do not share a common neurobiological mechanism.

Multiplexed memories: a view from human cortex

Authors:

• Robert T. Knight

• Howard Eichenbaum

Date: 2013

DOI: 10.1038/nn.3341

PubMed: 23434977

View PDF

Abstract:

A study recording directly from the human brain shows that connectivity between the prefrontal cortex, parietal cortex and the medial temporal lobe across different frequency bands underlies successful memory retrieval.

Age-related frontal-parietal changes during the control of bottom-up and top-down attention: an ERP study

Authors:

• Ling Li

• Caterina Gratton

• Monica Fabiani

• Robert T. Knight

Date: 2013

DOI: 10.1016

PubMed: 22459599

View PDF

Abstract:

We investigated age-related changes in frontal and parietal scalp event-related potential (ERP) activity during bottom-up and top-down attention. Younger and older participants were presented with arrays constructed to induce either automatic "pop-out" (bottom-up) or effortful "search" (top-down) behavior. Reaction times (RTs) increased and accuracy decreased with age, with a greater age-related decline in accuracy for the search than for the pop-out condition. The latency of the P300 elicited by the visual search array was shorter in both conditions in the younger than in the older adults. Pop-out target detection was associated with greater activity at parietal than at prefrontal locations in younger participants and with a more equipotential prefrontal-parietal distribution in older adults. Search target detection was associated with greater activity at prefrontal than at parietal locations in older relative to younger participants. Thus, aging was associated with a more prefrontal P300 scalp distribution during the control of bottom-up and top-down attention. Early latency extrastriate potentials were enhanced and N2-posterior-contralateral (N2pc) was reduced in the older group, supporting the idea that the frontal enhancements may be due to a compensation for disinhibition and distraction in the older adults. Taken together these findings provide evidence that younger and older adults recruit different frontal-parietal networks during top-down and bottom-up attention, with older adults increasing their recruitment of a more frontally distributed network in both of these types of attention. This work is in accord with previous neuroimaging findings suggesting that older adults recruit more frontal activity in the service of a variety of tasks than younger adults.

Oscillations and behavior: the role of phase-amplitude coupling in cognition

Authors:

• Boaz Sadeh

• Sara Szczepanski

• Robert T. Knight

Date: 2013

Human cortical sensorimotor network underlying feedback control of vocal pitch

Authors:

• Edward F. Chang

• Caroline A. Niziolek

• Robert T. Knight

• Srikantan S. Nagarajan

• John F. Houde

Date: 2013

DOI: 10.1073/pnas.1216827110

PubMed: 23345447

View PDF

Abstract:

The control of vocalization is critically dependent on auditory feedback. Here, we determined the human peri-Sylvian speech network that mediates feedback control of pitch using direct cortical recordings. Subjects phonated while a real-time signal processor briefly perturbed their output pitch (speak condition). Subjects later heard the same recordings of their auditory feedback (listen condition). In posterior superior temporal gyrus, a proportion of sites had suppressed responses to normal feedback, whereas other spatially independent sites had enhanced responses to altered feedback. Behaviorally, speakers compensated for perturbations by changing their pitch. Single-trial analyses revealed that compensatory vocal changes were predicted by the magnitude of both auditory and subsequent ventral premotor responses to perturbations. Furthermore, sites whose responses to perturbation were enhanced in the speaking condition exhibited stronger correlations with behavior. This sensorimotor cortical network appears to underlie auditory feedback-based control of vocal pitch in humans.