Analysis of Electrocorticographic Signals

There are 2 separate experiments proposed, both of which use repeated-measures designs.

1: Electrocorticography (ECoG) study of visual working memory. Each trial from the behavioral
task will start by presenting subjects with two visual images, one each from two of these
three categories: faces, words, and outdoor scenes. They will then be cued as to which one
they'll be tested on with a recognition probe, and after the first probe the cuing-probing
process is repeated. Patients selected for this study will have depth electrodes implanted in
the left medial temporal lobe and/or grids covering left occipital, temporal, and/or parietal
cortex, and suitability of a patient's data for the final dataset will require that a minimum
of one stimulus category can be decoded from them. (The precise minimum number of trials
required cannot be calculated a priori, because this requires knowing the signal-to-noise
ratio in a dataset, a property that is highly variable in electrocorticography data.)

2. Electrocorticography of spatial selective attention. Each trial from the behavioral task
will start by presenting subjects with a white "+" on a screen, with each arm pointing to a
potential target location. During each 92-trial block of trials, only two 180-degree opposing
locations will ever be cued, with one arm of the "+" turning yellow and the opposing one
turning blue, to indicate with 75% validity the location at which an oriented Gabor patch
will appear (5 degrees from fixation; cue color mapping counterbalanced), requiring a speeded
"R/L" tilt judgment. Orthogonal to cue-color configuration, half of the trials in each block
will begin with presentation of an "x" that will rotate by 45 degrees with an unpredictable
lag (.5 sec +/- .3). On these trials, the cue-to-target interval (i.e., from rotation to "+"
to color-cue onset) will be 750 msec. On trials that begin with the onset of a "+",
cue-to-target interval will vary unpredictably between 650, 750, and 850 msec. Decomposition
of alpha-band oscillations (brain waves cycling at roughly 10 times per second) into
components associated with each location will be derived by filtering the whole-scalp signal
with weights from the inverted encoding model trained to encode the four critical locations.

Inclusion Criteria:

- Participants with implanted electrode arrays who are willing to participate and able
to cooperate and follow research instructions will be recruited.

- Must be able to read

- Must be able to name objects

- Must be able to articulate thoughts with spoken language

Exclusion Criteria:

- post-operative pain requiring narcotics

- repeated seizures clouding consciousness

- IQ of 85 and below

- post-operative subdural bleeding

- cerebral pathology affecting the cortical regions from which recordings are made

- women who are pregnant, or who think they may be pregnant