Shedding light on brain function: the event-related optical signal

1
Shedding light on brain function the
event-related optical signal

• Gratton Fabiani (2001)

2
Functional neuroimaging

• Hemodynamic techniques
• PET and fMRI useful for spatial information about
  neural activity
• Electromagnetic techniques
• EEG and MEG useful for temporal information about
  neural activity

3
Event-Related Optical Signal

• Hemodynamic techniques lack temporal specificity
  and electromagnetic techniques lack spatial
  specificity.
• EROS provides both, good temporal and spatial
  information.

4
How does EROS work?

• Fiber optic cables act as sources and detectors.
• Sources have just one fiber, detectors have many.
• Sources carry light from lasers or LEDs.
• Photomotopliers function as detectors of photons.

5
How does EROS work?
6
Basic principles in the physics of optical
imaging.

• If photons are emitted from a light-source (fiber
  optic) against the surface of a semi-infinite,
  homogenous object they can be modeled using the
  same equations as those that describe the
  positive half of a dipole.
• Depth in the case of EROS depends partially on
  the distance between the source and detector.

7
How does EROS work?

• As light propagates from the source it gets
  scattered and absorbed by brain tissue.
• Changes in the activity of brain tissue affect
  the amount of scattering and absorption.
• Scattering causes the photons to have a longer
  transit time from source to detector.
• Therefore, scattering (activity) can be estimated
  from the increase in transit time/phase delay.

8
Phase delay
Input
Output
Intensity
Time
9
Experimental Support

• Visual stimulation experiment shows transit time
  increase for activated areas mm apart.

10
Experimental Support

• Comparison found good temporal and spatial
  overlap with ERPs and fMRI.

11
What might explain this response?

• Changes in neuronal membrane affect transparency
  of membrane and diffraction.

12
Overview neuronal activity

• Neuronal membranes distend/shrink as ions H2O
  move in/out after an action potential.

13
The Event-Related Optical Signal

• Pros
• Good temporal resolution (milliseconds)
• Good spatial resolution (lt1cm)
• Useful for studying neurovascular coupling
• Cons
• Penetration from scalp is limited to 3-5cm
  (cortex)
• Low signal-to-noise ratio requires averaging
• Mariannas question 1
• Pulse correction, phase rejection, movement
  artifact

14

• CNL EROS Video

15
Imaging cortical dynamics of language processing
with the event-related optical signal

• Tse, Lee, Sullivan, Garnsey, Dell, Fabiani and
  Gratton (2007)

16
Motivation

• How do the temporal cortex and inferior frontal
  cortex interact?
• fMRI is too slow to view this interaction
• Does processing differ for syntactic vs. semantic
  anomaly?
• Can EROS image interactions between cortical
  areas?

17
Hagoort (2005)

• Identifies three functional components
• Memory store of language information, involved
  in retrieval
• Unification integration of lexically retrieved
  information
• Control language to action, such as choose
  between using one of two languages

C
U
M
18
Hagoort (2005)

• Increased response in left inferior frontal when
  unification load is increased (in response to
  anomalous critical word). A lesser response also
  occurs for correct sentences.

19
Hagoort (2005)

• Model emphasizes that posterior and dorsal areas
  integrate syntactic information, while anterior
  and ventral areas function more for semantic
  integration.
• But, there is a lot of overlap.
• Preview of Alberts question
• Little evidence extending this functional
  specialization to the temporal lobe.

20
Methods Participants

• 16 participants
• All right-handed native English speakers
• 11 females and 5 females, ages 18-30
• Lucys question
• Why the disproportionate of females?
• Control for sinistrality?

21
Methods Materials

• Each participant saw 864 sentences
• 336 unacceptable sentences
• 144 were semantically anomalous at the final word
• The hungry child ate the floor.
• 48 filler sentences contained semantically
  anomalous words in medial position to prevent
  expectations that anomalies only occur in the
  final position
• 96 had grammatical violations of subject verb
  agreement
• If work isnt done, it pile
• 48 had grammatically incorrect pronoun case
• My mother promised to buy I

22
Methods Materials

• 528 acceptable sentences
• 144 controls for the semantic sentences
• 96 for syntactic subject verb agreement
• 48 for pronoun case anomaly
• 240 sentences to ensure subjects expect most
  sentences to be acceptable
• Length and frequency was accounted for across
  conditions.
• Israels questions-
• Couldnt syntactic anomaly be sentence-final?
  (When it rains, it pour(s)
• How long is the experiment?
• Approx. 10 hours total broken up into two 5-hour
  sessions.

23
Materials Procedures

• Sentences were presented word-by-word at the
  center of screen and subjects had to judge if
  sentence was well-formed.

24
ERP recording

• ERPs were recorded.
• EEG recording used four scalp electrodes (Fz, Cz,
  Pz, and right mastoid).
• Final bandpass filter of 0.1-20Hz
• Sampled at 100Hz
• Time-locked 1500 ms epochs with 200ms
  pre-stimulus onset

25
EROS recording

• EROS recording used two montages.
• Laser diodes emitted 830nm light at 220MHz which
  was picked up by PMTs modulated at 220MHz
• Sources, detectors, nasion, preauricular points
  and random points were digitally localized using
  a Fastrak 3D digitizer.
• Coregistration with individual subjects anatomy
  provided by MRI.

26
EROS setup

• 128 source-detector pairs per montage

27
Result Behavioral

• Most sentences were classified correctly
• Semantically acceptable- 94
• Semantically unacceptable- 95
• Syntactically unacceptable 86
• Syntactically acceptable- 88
• Analyses were performed only on these correct
  trials.

28
Results ERP

• Semantic unacceptable acceptable
• Difference waveform computed from 200-500ms
  peaking at 420ms (plt.001)

29
Results ERP

• Syntactically unacceptable acceptable
• Difference waveform computed from 500-1500
  peaking at 860 ms (plt.001)

30
Results EROS

• Significant increase in phase delay for anomalous
  critical words.
• Both conditions elicited S/MTC activation
  followed by IFC activation.
• Pattern occurred a few times for semantic
  condition suggesting oscillatory behavior.
• ROIs analyzed independently

31
Results EROS
32
ResultsEROS
33
Results EROS

• Different areas activated for each condition up
  to 665ms.
• Semantic ventral anterior/middle
• Syntactic dorsal posterior temporal
• More frontal activation for semantic anomaly, but
  lots of overlap.
• EROS signals predicted the N400 (_at_179ms, 384ms in
  S/MTC) and P600 (_at_ 819ms, 914ms in IFC)in
  semantic and syntactic condition, respectively,
  but not vice versa.
• Double dissociation in EROS/ERP

34
Results EROS

• Contrasts were 17mm apart along inferior
  -superior, but not anterior-posterior
• After 655ms there was no reliable differences
  between the activity for contrasts.
• Alberts question
• Potentially due to response
• Hagoorts model

35
Discussion

• EROS successfully showed interaction between
  temporal-frontal network involved in language
  processing.
• Supports model in which retrieval occurs in the
  temporal regions and the integration occurs in
  the frontal regions.

36
Discussion

• Might reflect integration process in IFC in which
  predictions about upcoming words are generated
  and sent to temporal areas.
• Mariannas question 1
• Lucys question 1 2
• Lynns Question
• Since syntactic anomalies were relatively easy to
  rectify there was little frontal activity.
• Semantic anomaly more difficult to correct and
  hence more back and forth.
• More extensive frontal activity
• Double checks retrieval?
• The location of these networks is consistent with
  previous fMRI studies.

37
Word position

• Syntactic words were in medial position while
  semantic were final.
• Is activation comparable for sentence medial and
  sentence final positions?

38
Word position
Effect is still present just smaller, as is
usually the case.
39
Word position