of 30
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Modeling Neural Network Connectivity at Rest and During Tasks with fMRI and ERP data: An Overview

Category:

Documents

Publish on:

Views: 15 | Pages: 30

Extension: PDF | Download: 0

Share
Description
Modeling Neural Network Connectivity at Rest and During Tasks with fMRI and ERP data: An Overview. Donald A. Robin, PhD Rachel Hutson , BA Chief, Human performance Division Research Imaging Institute Professor Neurology, Radiology, Biomedical Engineering
Transcript
Modeling Neural Network Connectivity at Rest and During Tasks with fMRI and ERP data: An OverviewDonald A. Robin, PhDRachel Hutson, BAChief, Human performance DivisionResearch Imaging InstituteProfessorNeurology, Radiology, Biomedical EngineeringUniversity of Texas health science center at san AntonioBrain ConnectivityAnatomicalFunctional EffectiveMeta-analytic Connectivity Modeling (MACM)Activation Likelihood Estimation (ALE)Reported foci of activation are treated as spatial probability distributions centered at given coordinates.Permutation tests are calculated to obtain an ALE null-distribution to identify convergence of foci across studies differentiated from random clustering (limited to gray matter). The histogram of the ALE scores following permutations is used to assign P values. False Discovery Rate and minimum volume (mm3) thresholds are applied to determine statistical significance.Inference reflects a null-hypothesis of random spatial association between experiments (i.e., random effects).GoalsTo develop an ontological system of describing function–structure correspondencesTo develop a probabilistic atlas of function-structure correspondenceTo develop a probabilistic atlas of inter-regional correspondenceInclusion Criterion A study is included in BrainMap if:It is published in a peer-reviewed, English-language journal Reports peaks of activation in MNI or Talaraich coordinates Reports whole-brain analysesCoding SchemeDescription of subjectsExperimental conditionsStimulus input Response modalityInstructionsCoded Meta-dataContext: purpose of the work (e.g., normal mapping, age effects, disease effects, drug effects)Behavioral Domain: neural systems studied: cognition, action, perception, emotion, interoception, or pharmacology.Paradigm class: challenege presented – e.g., anti-saccades, Stroop, reward-based learning.ROIsResting State fMRIUncovers intrinsic brain networks associated with specific systems (e.g., motor, attention) and a more general Default Mode Network that deactivates during taskLow frequency fluctuations in BOLD signal that are coherent in networks associated with a given functionResting State fMRI
  • Represent a relatively permanent brain state
  • Unbiased by task
  • Extremely robust - present during sleep and even anesthesia, present in non-human primates and perhaps non-primates
  • Connectivity is related to a given state or subject characteristic (e.g., perceptual measures of voice)
  • Resting State fMRI
  • High Clinical Potential
  • Subjects simply lies still in the scanner with eyes closed
  • Sequence <10 minutes (ours is 8)
  • Sensitive to disease/pathology
  • Network connectivity is known to systematically change (are malleable, show neuroplasticity) with behavioral training or treatment
  • Two Analysis Approaches to RSNRegion of interest (ROI) or Independent Component Analysis (ICA)In part, choice depends on questions and if a theory or existing data can drive ROI selectionICA preferred if no driving theory for exampleWe used a ROI seed based approach as we knew what areas would isolate vocalization network from general speech networkDefault Mode Network (DMN)Identified using ICA of resting-state dataNumerous disorders have alterations in the DMNMalaak N. Moussa, et al., 2012.DCM (Bayesian Models)Generative models of brain responses that provide posteriorestimates of effective strength of synaptic connections and their context dependencyDCM (Bayesian Models)Dynamic using differential equations for describing (hidden) neuronal dynamicsCausal (control theory) as they describe how dynamics in one population cause dynamics in another and how these interactions are modulated by experimental manipulation or endogenous brain activityDCMsDCMs strive for neurophysiological interpretabilityUse of a biophysically motivated and parameterized forward model to link modeled neuronal dynamics to measured dataAre Bayesian, thus predictive; each parameter is constrained by a prior distributionExamples of DCM in ERP studiesHypothesis abouta neural systemThe DCM cycleStatistical test on parameters of optimal modelDefinition of DCMs as systemmodelsBayesian modelselection of optimal DCMDesign a study thatallows to investigatethat systemParameter estimationfor all DCMs consideredData acquisitionExtraction of time seriesfrom SPMsDCM for ERP
  • DCM estimates effective connectivity
  • The influence one neuronal system has over another
  • DCM for ERP estimates effective connectivity by inverting a spatiotemporal forward model of the observed EEG data
  • In this study we used dynamic causal modeling (DCM) to model effective connectivity of ERP responses to pitch perturbation in voice auditory feedback in musicians with relative pitch (RP), absolute pitch and non-musician controls.
  • Effective Connectivity Associated with Auditory Error Detection in Musicians with Absolute Pitch(Parkinson et al, 2014, Front. Neurosci. PMID 24634644)Pitch-shift ParadigmSubjects listen to their own voice being played back in headphones while vocalizingUnexpected shifts are introduced to the feedback and subjects respond by shifting their own voice in the opposite direction to the shiftThis response is known as the pitch-shift reflexUpward pitch-shiftBlack lines: controlRed lines: voice F0 +30 cents Pitch–shift methods11 subjects were recruited to each of the absolute pitch (AP), relative pitch (RP) and non-musician (NM) groups.Subjects sustained vowel /a/ phonation for 2 second intervals± 100 cent shift magnitude 200ms shift durationPresented 500-1000ms after voice onset32 channel ERP recordingVoiceAuditory feedback of the voicePitch shift during vocalizationDCM Model Specification
  • We used Bayesian model selection (BMS) family level inference to examine the following 2 factors in all three groups:
  • Factor 1: Effect of STG modulation across hemisphere
  • Factor 2: Effect of bilateral, left or right hemisphere modulation of connections
  • Significance of coupling parameters was also directly compared across all groups for all modulated connections of the bilateral family of models.
  • We identified a basic model, including modulated connections from STG to PM, PM to STG and STG to IFG in both hemispheres based on fMRI activation during vocalization
  • DCM ModelsFactor 1: Effect of STG modulation across hemisphereRight to LeftBilateralLeft to RightBoth hemispheresFactor 2: Effect of bilateral, left or right connectionsLeft hemisphereRight hemisphereERP ResponsesData were modeled between 1 and 200ms following the pitch-shift stimulusModel Selection ResultsFactor 1 – Effect of STG modulation across hemisphere, no significant families identified for each groupalthough the AP group clearly favored models with modulated left to right STG connectionsModel Selection ResultsFactor 2 – Effect of bilateral, left or right modulation of connectionsBMS indicated that modulation of STG connections to PM and IFG in both hemispheres is critical in the identification of self-voice pitch error in musicians with AP (0.92 random effects model exceedance probability) but not in the RP and NM groups (Figure 3).Coupling ParametersInfluence on couplingNegative coupling of left hemisphere PM to STG connections in AP and RP groups, compared to a positive coupling in the NM group was identifiedConclusionsMusicians with enhanced pitch discriminating abilities likely have a “fine tuned” auditory error detection and correction system involving modulation of left to right STG connections.The role of the left hemisphere is radically different in musicians than in non-musicians. We identify reduced connectivity of left hemisphere PM to STG connections in AP and RP groups during the error detection and corrections process relative to non-musicians.This suppression may allow for enhanced connectivity relating to pitch identification in the right hemisphere in those with more precise pitch matching abilities Our findings here also suggest that individuals with AP are more adept at using feedback related to pitch from the right hemisphere.AcknowledgmentsNIH 1R01DC006243The Robin LabAmy Parkinson, PhD Annie NewRachel SmallwoodRachel HutsonMichael WaskiewiczAurora RobledoCollaboratorsCharles Larson, PhDJeremy Greenlee, MDRoozbehBehroozmand, PhDSona Patel, PhD
    Similar documents
    View more...
    Search Related
    We Need Your Support
    Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

    Thanks to everyone for your continued support.

    No, Thanks