Magnetoencephalography

Papanicolaou 1998 Fundamentals of Functional Brain Imaging

Functional Brain Imaging

• The brain is constantly sending electrochemical signals

• Coordinated signaling activity of a large set of neurons somewhere in the brain creates a transient deviation in electromagnetic signal intensity beyond the normal range of variation

• Temporal variations of the signal can be associated with functions

Measuring brain activity

• Neurons are the current sources – Synaptic currents– Dendritic currents– Axonal currents

• These currents are very small• Neurons must be in a open field

configuration in order for the combination of sources to be strong enough to be measured

What are we measuring?

• Synaptic and axonal currents cancel each other out

• Dendritic currents are the primary source of electrical activity that can be measured

• Two electromagnetic signals arise from the primary source– Secondary (volume) electrical currents– Magnetic flux (magnetic fields)

Volume Currents

Electricity

• Volume currents follow the path of least resistance in the extracellular components

• EEG measures the volume currents at the surface of the head

• Volume currents become distorted and attenuated

• Overcoming this distortion was one major factor in the development of MEG

Properties of magnetic fields

• An electric current will induce a magnetic field perpendicular to its direction– A magnetic field will induce a current

• Magnetic fields are not distorted as they emerge from the brain source

• Magnetic field strength:– is proportional to the strength of the source current

(amperes).– dissipates as a function of the square of the distance

from the current source.

Magnetometer

• Loop of wire at the head surface• Magnetic field at the surface of the head

creates a current in the wire by induction– Only the portion which is perpendicular to the

head

• Strength of the current is proportional to the strength of the magnetic field

• The surface distribution can be determined with enough magnetometers

Amplification

• The induced currents are very small– Smaller than thermal noise

of conventional amplifiers

• Superconductive Quantum Interference Devices (SQUIDS)– Wires are cooled to 4 K to

lower the resistance and housed in a thermally insulated drum (a dewar)

Dipolar Distributions

• If you have a current line in a sphere (e.g. a head), the magnetic field will create a dipolar distribution

• On the surface of the sphere, only the perpendicular part of the vector can be measured

• Two extrema points are where the field strength is the highest

Forward Problem

• Calculating an effect or predicting a phenomenon from a set of known causes and antecedent conditions.

• If we know the cause (the sources) we can uniquely predict the effect (the magnetic distribution) using the Biot-Savart Law– Strength, orientation, (x, y, z) location– Predicts surface distribution

Inverse Problem

• If we do not know the cause (or the sources) we cannot uniquely determine the effect (or the distribution).

• We can hypothesize what the sources are and determine a hypothesized distribution

• We iteratively change the parameters until we are satisfied that the hypothesized distribution matches the actual distribution.

Dipolar Distributions

• Source location must be below the midpoint

• Source location must be at a depth proportional to the distance between the extrema

• Source strength related to absolute intensity of flux (at a given depth)

• Source orientation related to orientation of extrema

Coordinate system

• Y-axis: line between pre-auricular points

• X-axis: line between nasion and perpendicular intersection of midpoint of the y-axis

• Z-axis: line perpendicular to x-axis and y-axis at the midpoint

• Labeled with lipid markers and structural MRI is taken.

MEG Validity

• How sure are we that MEG images are accurate?

• Consistent when compared with lesions, intracranial electrophysiology, and behavioral tasks

• Limited spatial resoultion

• Excellent temporal resolution

Neural dynamics of reading morphologically complex words

Vartiainen et al. 2009 NeuroImage

Research Question

• How are morphologically complex word forms represented and processed in the brain?– “book+s”

• ‘books’

– “koulu+i+ssa+mme+kin”• ‘even in our schools’

• Do inflected words require additional processing early or late in the assumed sequence of cognitive operations?

Traditional Views

• Two separate lexical entries for book and books– Butterworth, 1983

• Decomposition into constituent morphemes– Taft and Forster, 1975

• Dual-route mechanism where low-frequency words are processed through decomposition and high-frequency words are processed in the full-form– Chialant and Caramazza, 1995

Other Factors

• Frequency– Alegre and Gordon, 1999

• Type of morphological operation– Miceli and Caramazza, 1988

• Inflectional complexity of language– Lehtonen et al., 2006a

• Language background of speaker– Portin et al., 2008

Inflectional processing

• Word form level– Decomposition of stem and affixes

• Lemma level– Lexical representations of stem and affixes

are recombined– Evidence indicates that the increased

processing is during the semantic-syntactic level

• Pawel’s question

Lexical, Syntactic, or Semantic?

• Hyönä, Vainio, and Laine 2002 European Journal of Cognitive Psychology

• A lexical decision task with isolated words showed more effortful processing for inflected than monomorphemic nouns

• However, this morphological complexity effect did not generalise to words within sentence contexts– Fixation durations– Response latencies

Finnish

• Finnish is a morphologically complex language– Effects of this complexity should be evident

• Inflected Finnish nouns elicit longer reaction times and/or higher error rates when compared with otherwise matched monomorphemic nouns

• There is a difference between inflected words and noninflected words in high frequency words, less so in middle frequency words, and non at all in low frequency words– Lehtonen and Laine, 2003

Cognitive processing

• Lynn’s question• Lyytinen 1987 Scand J

Psychol– Correlated cognitive

components with inflectional errors

– Block classification and integration skill of visual-linguistic information

– Significance of memory increased as a function of the relative linguistic component

Hemodynamic studies of readingFunction Area of the brain

Word Left occipito-temporal cortex

Fusiform gyrus

Lemma Left posterior middle temporal gyrus

Left basal temporal area

Left inferior frontal gyrus

Syntax Left middle temporal gyrus

Left temporal pole

Left posterior superior temporal sulcus

Left inferior and middle frontal gyrus

Hemodynamic Localization

• Effects were typically reported in activation of the left inferior frontal gyrus, interpreted to reflect analysis of grammatical features, and the left temporal regions, thought to denote access to the semantic representations of the stem and affix

MEG EvidencePylkkänen and Marantz, 2003

Salmelin, 2007

• Basic visual analysis– Midline occipital cortex– 100 ms

• Letter-string analysis– Left occipito-temporal cortex (fusiform gyrus)– 140-200 ms

• Lexical-semantic activation– Left superior temporal cortex– 200 ms to 600-800 ms

• Fiorentino and Poeppel 2007 (~350 ms, English)

Background

• EEG: Lehtonen et al., 2007– 450 ms onwards– N400 and positive component larger for

inflected words

• fMRI: Lehtonen et al., 2006b– Left posterior superior temporal sulcus– Left inferior frontal gyrus

Subjects

• 5 males• 5 females• Mean age 30• Age range: 25-46• Normal vision (7)• Corrected-to-Normal vision (3)• Lucy’s question

– Lack of information about subjects

Stimuli

• 22.7 million word newspaper corpus

• Analyzed using WordMill Lexical Search

• The frequency range was allowed to be broader for the high-frequency range (26.4-504 occurrences per million) than the low-frequency range (0.04-4.23) to find a large enough number of high-frequency monomorphemic items of sufficient length

Stimuli

• Syntactic case (75%)– Genitive– Partitive

• Marianna’s question– Frequency of inflectional

morpheme– Effect of idioms on

inflectional morphemes

• Semantic case (25%)– Inessive– Elative– Illative– Adessive– Ablative– Allative

• Related case– Essive

Stimuli

• 320 stimuli in four groups of 80

Procedure

• Words were presented visually one at a time• The MEG response timing was corrected for the

34-ms delay from stimulus trigger• Each noun was shown for 400 ms, and the

stimulus onset asynchrony was 3000 ms• Stimuli from the different categories were

presented in a random order and divided into 4 blocks

• Subjects were instructed to read the words silently

Procedure

• 20 additional target words were presented randomly– 5 words per category were matched similarly to the actual stimuli

• The word was followed by a question mark– 1500 ms after the word onset– Duration 1500 ms– Prompted the subject to read out loud the preceding word

• The next trial started after a delay of 1500 ms• The target trials as well as the trials immediately

following the question marks were not included in the analysis.

• This is one main difference from previous studies

MEG

• 102 triple sensor elements

• Two orthogonal planar gradiometers

• One magnetometer

• Band-pass filter: 0.03 and 200 Hz

• Digitized at 600 Hz

• Averaged: 0.2 s to 1 s after stimulus

MEG

• Epoch rejected if EOG signal > 150 μV

• Average 67 artifact-free trials from 80 trials

• Israel’s question– Rejecting artifact

Epoch rejection

Areal mean signal (AMS)

• Left frontal• Right frontal• Left temporal• Right temporal• Parietal

• Left rolandic• Right rolandic• Left occipito-temporal• Right occipito-

temporal• Occipital

Time windows

• 50-170 ms

• 170-330 ms

• 330-500 ms

Equivalent Current Dipole

• ECDs identified individually• ECD location and orientation fixed while

amplitude varied• Head coordinate system set by nasion and ear

canals– Head Position Indicator coils

• Maximum peaks during the time windows were compared (repeated measures ANOVA)– Word frequency– Morphological complexity

Results

• The response was stronger to inflected than to monomorphemic words at 330–500 ms– Left temporal region– Parietal region

• The response was stronger to the low- than high-frequency words at 330–500 ms– Left temporal region

• The response was stronger to the low- than high-frequency words at 170–330ms– the parietal region

• There were no significant effects in the 50–170 ms time window over any region

Results

• ECDs from individual subjects, each representing the center of an active cortical region, were grouped according to similarity in location and time course of activation

• Duplicated results from previous studies

MEG EvidencePylkkänen and Marantz, 2003

Salmelin, 2007

• Basic visual analysis– Midline occipital cortex– 100 ms

• Letter-string analysis– Left occipito-temporal cortex (fusiform gyrus)– 140-200 ms

• Lexical-semantic activation– Left superior temporal cortex– 200 ms to 600-800 ms

Results

Results

• Inflected low-frequency words– Strongest activation

• Monomorphemic high-frequency words.– Weakest activation

Results

• Inflected low-frequency words– Longest duration (35

ms)

• Monomorphemic high-frequency words.– Shortest duration (10

ms)

Results

• No significant interaction was found between Morphological complexity and Word frequency in any of the measures.

MEG fMRI/PET

• No activation in frontal cortex

• No activation in Visual Word Form Area

• Rapid changes of highly synchronized neural activation

• Silent reading task

• Activation in inferior frontal cortex

• Activation of VWFA• Changes in blood

flow, oxygenation, and glucose uptake

• Lexical decision task

Question

• The neural response was stronger and longer-lasting to the inflected than to the monomorphemic words, suggesting decomposition of all the inflected words throughout the frequency range used in the study.

• However, there was no statistically significant interaction between morphology and word frequency, which implies that morphological decomposition occurred for inflected words throughout the frequency range employed.

Conclusions

• All inflected words in Finnish are decomposed• Very high frequency inflected words may be an

exception to this rule• Only those inflected words that are of very high

frequency in the Finnish language may acquire full-form representations.– Laine et al., 1994– Niemi et al., 1994– Soveri et al., 2007

Conclusions

• Morphological effects in the left temporal cortex beginning around 200 ms after the word onset

• Reliable effects of Finnish morphology have not been detected with any imaging modality in early time windows or in the regions assumed to be involved in pre-lexical analysis of written words

• Inflectional processing cost stems from the semantic–syntactic level