Neural Correlates of Human Cognitive Function: A Comparison of Electrophysiological and Other Neuroimaging Approaches Leun J. Otten Institute of Cognitive Neuroscience & Department of Psychology University College London
Overview of Lecture Functional neuroimaging: the aim of the game Methods When to use what? Strengths and weaknesses Integration of imaging modalities Conclusion: Why ERPs? Interpreting ERPs
Functional Neuroimaging An approach in cognitive neuroscience to understand: the identity and organization of human cognitive functions how these are instantiated in the brain Cognitive Functions Neural Activity A B C D
How the Brain Works? Processing More Processing Even More Processing Time Region 1 Region 2 Region 3 early sensory late higher cognitive
Onset Latency of Visually Responsive Neurons All regions active within 100 ms From Robinson & Rugg (1988).
Onset of Stimulus Selective Activity in Monkey Prefrontal Cortex Activity related to stimulus discrimination also observed within 100 ms From Rainer et al (1988).
How the Brain Really Works? Region 1 Region 2 Region 3 early late Time Region 1 Region 1 Region 1 Region 2 Region 2 Region 2 Region 3 Region 3 Region 3
Aim of the Game Must be able to localize as well as identify the temporal organization of neural activity associated with cognitive functions.
Neuroscience Methods PET CT Diffusion imaging EEG Lesions Single cell recordings TMS Etc _ fmri _ Computational modeling _ SPECT _ MEG _ Genetic manipulations _ Multiple cell recordings _ Optical imaging _ Etc
Functional Neuroimaging Methods Noninvasive measurement of brain activity in humans Two classes: Haemodynamic - brain blood supply Electrophysiological - electromagnetic fields
Haemodynamic Methods Close coupling between neural activity and blood supply Positron Emission Tomography (PET): measures regional cerebral blood flow (rcbf) radiotracer introduced in blood stream accumulates in brain region proportional to amount of blood flow Functional Magnetic Resonance Imaging (fmri): measures Blood Oxygenation Level Dependent (BOLD) signal blood draining from region more oxygenated when region is active.
Electrophysiological Methods Activity from some neuronal populations detectable on scalp If neurons in population are oriented in same direction and activated synchronously, they create an electromagnetic field open field closed field Electrical component: Electroencephalogram (EEG) Magnetic component: Magnetoencephalogram (MEG) Usually, time-locked measures Event-related potentials (ERPs) or oscillations Event-related magnetic fields (ERFs) or oscillations
Event-Related Potentials/Fields Small changes in the EEG/MEG, time-locked to some event Recorded from several sites on the scalp Real-time representation of how an event is processed + Condition A Condition B 10µV 0 800 ms
EEG vs MEG EEG sensitive to both radial and tangential sources, MEG only to tangential sources MEG easier to localize because less spatial smearing, fewer distortions, and reference-free MEG easier to obtain high density recordings, but sensitive to noise EEG is CHEAP!
What Type of Inferences Can be Made? EEG/MEG fmri/pet 1. Identify that a function is engaged? 2. Identify distinct functions? 3. Identify temporal organization? 4. Identify neural substrate of function? YES YES YES YES YES NO NO YES
Strengths and Weaknesses EEG/MEG fmri/pet 1. Measure of neural activity direct indirect 2. Temporal resolution high low 3. Spatial resolution low high 4. Design type blocked/ blocked/ event-related event-related 5. Sampling of regions heterogeneous (near) homogeneous 6. Sampling of neural activity* partial partial Spatial Resolution (mm) 100 10 1 EEG fmri 0.01 0.1 1 10 Temporal Resolution (s)
Preconditions for Detecting a Signal ELECTROPHYSIOLOGICAL Activation must be synchronous Neurons must have an open field configuration BUT Activity need not be extended in time Signal sensitive to changes in both timing and amplitude of activity HAEMODYNAMIC Activity need not be synchronous Orientation of activated neurons irrelevant BUT Changes in neural activity detectable only with change in metabolic demand Signal amplitude influenced by magnitude and duration of neural activity*
Scaling of Haemodynamic Response with Duration of Neural Activity amplitude 10 5 2 1 0.5 0.1 5 15 25 seconds
Consequences Two modalities do not necessarily reflect same neural activity Absence of effect does not necessarily mean that no differential brain activity occurred
Integrating EEG/MEG and PET/fMRI How? Informal: perform parallel experiments argue by analogy for equivalence of findings of interest Formal: perform same experiment on same subjects use formal source localization procedures to link data EEG/MEG ill-posed inverse problem Anatomical constraints Reconstructed sources Functional constraints
Example: Informal Converging ERP/fMRI memory retrieval effects? 500-800 ms 1100-1500 ms
Example: Formal Integrating fmri and MEG data Dale et al. (2000). Neuron, 26, 55-67.
Conclusion: Why ERPs? Allows to identify when cognitive functions operate dynamics of cognitive functions Cheap and easy method to fractionate cognitive functions In combination with other modalities, allows to identify functional brain networks
How to Interpret ERPs? ERP Effect Yes Known? No Use as marker of process associated with the effect Scalp distribution: Which process Time course: When process is active Amplitude: How active process is
Inferences: ERP Effect Unknown Site 1 A Condition 1 Condition 2 B Site 2 Site 3 + _ 0 500 1000 ms 0 500 1000 ms ERPs: 1. differ between conditions Cognitive processes differ between conditions 2. differ between conditions at time x Cognitive processes differ by time x 3. differ in distribution across the scalp Cognitive processes differ qualitatively. From Otten & Rugg (in press).
Inferences: ERP Effect Known A Condition 1 Condition 2 B Site 1 Site 2 Site 3 X' X + _ 0 500 1000 ms 0 500 1000 ms ERPs: Differ with respect to certain component X Cognitive process associated with component engaged to different degree or at different time From Otten & Rugg (in press).
ERP Components Components usually labelled by polarity and time/order of occurrence; sometimes given functional name Definitions: Theoretical: Signature on scalp of activity from specific neuroanatomical source(s), associated with specific functional process Operational: Part of the waveform with a circumscribed scalp distribution (alluding to the underlying neural configuration) and a circumscribed relationship to experimental variables (alluding to the cognitive function served by the activity of this configuration). Issues: Functional significance seldom completely understood, difficult to identify and/or dissociate components
Some Common Components ABR, P1, N1, P2 (early perceptual processing) P50, N1d, N2a, N2b, N2pc (attention and discrimination) P300/P3b/LPW, P3a, novelty P3 (context updating) LRP (motor processing) ERN (error processing) Left parietal, mid-frontal, right frontal effects (memory) N400, P600/SPS, LAN, N270, PMN (language) BP, CNV, NSW, PSW (preparatory processes) And many many more
Things to Keep in Mind (1 of 8) 1. ERPs contain not one, but many, effects ERPs represent neural activity that is aggregated across time and space Need to decompose waveforms.
Things to Keep in Mind (2 of 8) 2. Null result usually means little Condition 1 Condition 2 0 -> Time + _ Processes and neural activity not necessarily equal across conditions! Experiment may not have had enough statistical power Waveforms may not have been quantified or analyzed in the optimal way ERPs sample only a subset of total brain activity: activity only visible on the scalp when elements of a neuronal population activate or de-activate synchronously, and when their geometric configuration is such that their activity summates.
Things to Keep in Mind (3 of 8) 3. Meaning of scalp distribution differences unclear Different scalp distributions suggest different patterns of neural activity across conditions or time But, nature of difference unknown. Difference may be due to anatomically distinct generators, or different relative contributions (in terms either of their strengths or time courses) of different components of a common set of generators. A B A B vs
Things to Keep in Mind (4 of 8) 4. Polarity has no particular significance Polarity defines an ERP effect However, polarity determined by location of reference electrode, baseline against which the effect is compared, and location and orientation of intracerebral sources Polarity also varies because of neurophysiological reasons, e.g. whether input is inhibitory or excitatory and whether input is received via synapses distal or proximal to the cell bodies.
Things to Keep in Mind (5 of 8) 5. Scalp field intracerebral sources No transparent relationship between an electrical field observed on the scalp and the brain regions giving rise to that field Need formal source localization methods.
Things to Keep in Mind (6 of 8) 6. Amplitude strength of process Amplitude differences can occur in the absence of a change in the strength of the underlying neural activity Signal averaging assumes across-trial invariance. Amplitude differences can arise when: - Signals differ in time of occurrence across trials ( latency jitter ) and the degree of variability differs between conditions - Conditions have different proportions of trials carrying effects of constant amplitude. In that case, amplitude differences carry information about probability that process is engaged.
Things to Keep in Mind (7 of 8) 7. Time course is relative The onset of an effect does not necessarily reflect the actual point in time when the brain first distinguishes the conditions, but presents an upper bound Depending on the question of interest, time course may be characterized by various parameters including onset, latency to peak, rise time, and/or duration. Condition 1 Condition 2 Rise Time Duration Peak + _ Onset -> Time
Things to Keep in Mind (8 of 8) 8. Correlation vs causation All inferences from neuroimaging data are correlational Neural activity may not be necessary for a process, but may reflect processes downstream or incidental to it To study causation, must interfere with neural activity and study consequences Brain lesions, pharmacological manipulations, TMS.
Mapping Function onto Structure Must assume a one-to-one mapping between neural activity and cognitive processes Even then, distinct neural activities may not imply distinct functions The same process may act on different content, or rely on different neuroanatomical pathways (e.g. attention to left/right visual field) Distinct patterns of neural activity is a necessary, but not sufficient, condition to postulate distinct cognitive processes How different do two patterns have to be before they imply functionally distinct processes?.
Recommended Readings Handy, T. C. (Ed). Event-related potentials: a methods handbook. Cambridge, MA: MIT Press. (Nov 2004). Picton et al. (2000). Guidelines for using human event-related potentials to study cognition: recording standards and publication criteria. Psychophysiology, 37, 127-152. (Download from http://sprweb.org/~spr/journal.html) Chapters 1 and 2 of M. D. Rugg & M. G. H. Coles (Eds), Electrophysiology of Mind: Event-Related Brain Potentials and Cognition. New York, NY: Oxford University Press, 1995. Kutas, M. & Dale, A. (1997). Electrical and magnetic readings of mental functions. In M. D. Rugg (Ed), Cognitive Neuroscience. Hove: Psychology Press.