GONZALEZ, Alex, ECoG Parietal Memory Mechanisms, Wagner Lab, June 8, 2016
<p class="MsoNormal" >GONZALEZ, Alex, ECoG Parietal Memory Mechanisms, Wagner Lab, June 8, 2016</p>
<p>
this is a thesis defense (he is a PhD candidate in Anthony Wagner's Memory Lab)
Anthony does the v. humorous intro</p>
<p>
baby pic... see how the PhD has aged him.</p>
<p>
Alex does Tough Mudder runs... determination and stamina.</p>
<p>
ambitious and daring... new stuff! new tech!</p>
<p>
title is <b>Cortical Electrophys of Human Memory</b></p>
<p>
he went to watch a movie...</p>
<p>
went to watch <b>Jungle Book </b>(the NEW movie)</p>
<p>
he also saw it as a kid (the Disney version from 30 years ago)</p>
<p>
some new images triggered old images...from the old movie</p>
<p>
so... we call this episodic memory processes... how does it work?</p>
<p>
<b>encoding vs retreival</b>...</p>
<p>
<b>Encoding</b>... </p>
<p>
and did we recognize the old memory or was it a MISS (failure of retrieval)</p>
<p>
OUTLINE for today...</p>
<p>
retrieval... parietal lobe ECoG etc.</p>
<p>
fMRI (where in the brain) BOLD is indirect... ...
we get high spatial res but poor temp resol</p>
<p>
EEG (the WHEN) high temporal and direct measure of brain activ.</p>
<p>
but poor spatial res...</p>
<p>
ECoG get BOTH the when and the WHERE</p>
<p>
<b>ECoG electrocorticography...</b></p>
<p>
these are epilepsy pts = patients implanted with trodes pre-excision.</p>
<p>
direct e-phys of brain activity...</p>
<p>
high temporal and spatial res.</p>
<p>
shows 2 seconds worth of data... time varying btwn -100 and 100</p>
<p>
80 to 180 Hz band is tracking the population activity...</p>
<p>
this tracks neural firing just below the trode.</p>
<p>
take Hilbert transform of signal and can get envelope</p>
<p>
and then take log transform...</p>
<p>
</p>
<p>
Retrieval.. parietal lobe ECoG...</p>
<p>
did we see that bear before (from the movie) as a kid.</p>
<p>
(old) let's say, strong recognition</p>
<p>
vs Weak recogntion</p>
<p>
or OLD and we forgot about it</p>
<p>
vs NEW (ie, it was not there in the old movie)</p>
<p>
many refs.</p>
<p>
eg Shimamura 2014...</p>
<p>
we think MTL does encoding...</p>
<p>
coding of memory representations</p>
<p>
retrieval effects at 200 ms..</p>
<p>
now... parietal cortex...</p>
<p>
shows arrow going from occiput to the parietal
.. but also an arrow going from HC to parietal.</p>
<p>
what is timing in parietal cortex</p>
<p>
SPL = superior parietal lobule</p>
<p>
then IPS = intalparietal sulcus and below that is</p>
<p>
AG = angular gyrus...</p>
<p>
X axis high confidence that the stimulus is NEW
(not seen before) vs</p>
<p>
way to the left is high confidence that we have seen that stim before</p>
<p>
Y axis is activity in SPL...</p>
<p>
we call that decision uncertainty...</p>
<p>
shows a gaussian hump.</p>
<p>
VERSUS</p>
<p>
IPS tracks memory strength ... goes linearly from low to high</p>
<p>
(from NEW to (strongly confident) OLD.</p>
<p>
parietal cortex during memory retr... open Qs</p>
<p>
want a genl model of parietal lobe function....</p>
<p>
tempor dynamics in IPS vs SPL</p>
<p>
is SPL top down attn</p>
<p>
vs IPS accum of mem evidence towards decisions etc.</p>
<p>
shows trode coverage over parietal lobe...</p>
<p>
encoding task...</p>
<p>
1.5 to 3.5 sec fixation then</p>
<p>
latern and then few sec later the probe.</p>
<p>
eg grief then paper</p>
<p>
stim for one sec</p>
<p>
and </p>
<p>
stim comes on and then graph activity</p>
<p>
compares SPL vs IPS...</p>
<p>
this is gonzalez in PNAS in 2015...</p>
<p>
in SPL see divergence at around 500 msec after stim onset...</p>
<p>
btwn New vs Old...</p>
<p>
but with IPS get divergence earlier around 200 msec</p>
<p>
separ btwn Old vs New...</p>
<p>
shows an intial bump (with info entering and then old vs new later)</p>
<p>
now look at response locked data (when S responds)</p>
<p>
seens that activity in IPS has finished around 1 second...</p>
<p>
old vs new converge</p>
<p>
BUT in SPL... bigger spear all the way thru</p>
<p>
espec with def. new items...</p>
<p>
is it cuz this is more effortful...</p>
<p>
SPL shows higher activ at 600 ms for NEW items thru-out reponse.</p>
<p>
and seems that they is deffer. attn to novel items ie is THAT the explanation</p>
<p>
IPS</p>
<p>
acitivty is higher for OLD at 300 ms up to 200 ms before the response.</p>
<p>
consistent with acc tof accum of memory evidence...</p>
<p>
this rules out a pure visual field effect.
</p>
<p>
part 2...</p>
<p>
ENCODING</p>
<p>
see an image...</p>
<p>
eg Mogli the bear's lying in the river (old image)</p>
<p>
Strong memory is charac. by faster retr decison</p>
<p>
and weak memory has slower retr decison</p>
<p>
literature...</p>
<p>
brain activity reflects efficiency of encdoing</p>
<p>
MTL involved in binding of cortical reps</p>
<p>
and theta oscil</p>
<p>
PFC</p>
<p>
WM and selection of fetaures to be encoded.. modulated by theta</p>
<p>
parietal cortex..</p>
<p>
SPL and IPS allocation of top-down attn</p>
<p>
and perhaps using alpha oscil.</p>
<p>
*</p>
<p>
Encdoing Oscillations
fluctuations of ATTN modulates strength of encoding</p>
<p>
might be indep of stim onset.</p>
<p>
Alpha oscil is at abt 10 Hz.</p>
<p>
perhaps due to rhythmic fluctuation of inhib neurons...</p>
<p>
seeing more perhaps correl with disinhib</p>
<p>
and theta oscil at 6Hx</p>
<p>
are linked to LT memory formation</p>
<p>
oscil power as marker of cortical communic and computaiton..</p>
<p>
delta defined. as 1 to 4 Hz</p>
<p>
and theta as 4 to 8 Hz</p>
<p>
alpha 8 to 12</p>
<p>
beta 12 to 30 etc.</p>
<p>
and also low gamma (low freq) and
high gamma...</p>
<p>
relate this to task...</p>
<p>
doing ENCODING</p>
<p>
ITI = inter... interval... 1.5 to 3.5 sec</p>
<p>
say retr is slow (weak memory)</p>
<p>
relating data to behavior...</p>
<p>
how to relate activity. to fast vs slow decisons of S.</p>
<p>
many channels and then do PCA.. and relate to RTs = reactio times...</p>
<p>
shows correl btwn weak vs strong memories with one partic princ component</p>
<p>
and shows component spectrogram...</p>
<p>
get histogram of components...</p>
<p>
and only look at ones that have r > 0.15...</p>
<p>
channels (are the trode locations)</p>
<p>
by region the IPS is the location that has more correl with later memory.</p>
<p>
1st PC (prin comp)</p>
<p>
compres IPS vs SPL vs AG... all have sim patterns</p>
<p>
prob due to stim evoked... activ.</p>
<p>
but look at IPS 2nd comp.</p>
<p>
difft.</p>
<p>
and 3rd comp of IPS...</p>
<p>
we see less inhib and more theta power...</p>
<p>
perhaps more communic btwn HC and cortex... release for inhib.</p>
<p>
so, conclusion of Expt 2..</p>
<p>
IPS correlates with subseq memroy strength. more than do the surrounding regions</p>
<p>
and high-freq power coupled with lower few seen in all regions</p>
<p>
extracted patterns nt at multiplex of processes in thei region predict later memory...</p>
<p>
*
<b>3rd part of talk... electrical STIMULATION</b>...</p>
<p>
encoding... theta osc phase...</p>
<p>
is there a preferred encoding phase...</p>
<p>
we remember something more when it's at a peak of the oscil.</p>
<p>
ie, there is a stronger connex at peak of oscil... (this is at 6 Hz)</p>
<p>
at trough the connex are weaker...</p>
<p>
get LT potentiation at peak (from MTL/HC)</p>
<p>
and LT depr at the trough...</p>
<p>
hypothesis is that alignment is impt to remembering...</p>
<p>
he uses tACS = transcranial Alternat Currrent Stim...</p>
<p>
this induces an extra cell membrane voltage flucs</p>
<p>
uses Neuroelectrics Starsim...</p>
<p>
use contin theta 6Hz stim at FZ 1.2 mA)</p>
<p>
this is closed loop stimulation and it is locked to the recording of theta oscil...</p>
<p>
see Herman 2016...</p>
<p>
some tingling initially..</p>
<p>
shows image for 83 msec and repeat 6 times for total of one sec</p>
<p>
the ITI for 500 msec</p>
<p>
this is a face vs scene categ decision...</p>
<p>
so... encdoing... present stimulus at zero phase</p>
<p>
vs 144 degree out</p>
<p>
this is for encoding...</p>
<p>
and now shows dP d prime...</p>
<p>
scenes are better remembered than faces...</p>
<p>
and also probe for confidence...</p>
<p>
faces are harder to remember... lower confid...</p>
<p>
...</p>
<p>
tabulate later memory recog by encodoing stim phase...</p>
<p>
look at avg vector (to encode the phase in complex space)</p>
<p>
now, look at population alignment by confidence...</p>
<p>
remembered with high confi is seen at around - 135 degrees</p>
<p>
this is NOT random (would be uniform around the circle but it is NOT)</p>
<p>
remem high confi... phase is important</p>
<p>
ie alignment is max at + 225 or - 135.</p>
<p>
faces show more variance...</p>
<p>
concl of expt 3...</p>
<p>
subseq recognition mem phase aligns at around -135 degr.</p>
<p>
this is equv to 62 msec lag of the peak</p>
<p>
HC leads PFC by 50 msec</p>
<p>
(HC is where LTP occurs)</p>
<p>
alignment was driven by the Scenes stim</p>
<p>
ie scenes showed the highes tmemroy</p>
<p>
*</p>
<p>
Summary of everything...</p>
<p>
want to understand the timing of neuroanatom
processes neces. for how brain works...</p>
<p>
Expt 1
contrib to know of parietal lobe function</p>
<p>
expt 2</p>
<p>
evid for state-like act contrib to later memory</p>
<p>
expt 3</p>
<p>
modulate later recognition thru electrical stim.</p>
<p>
acknow...</p>
<p>
his comm..</p>
<p>
Chair.. EE fourier transforms...</p>
<p>
Julie psych</p>
<p>
Surya...</p>
<p>
Tony.. EEG master...</p>
<p>
main advisor is Anthony,,,</p>
<p>
demands perfection... pushes him to the max...</p>
<p>
balances family life...</p>
<p>
funding DARE Fellowship;</p>
<p>
Ford Fnd</p>
<p>
NIH BioTech Training..</p>
<p>
Jay Q timing of HC to parietal... about 300 to 400 msec... with recog...</p>
<p>
Kalanit Q diffs btwn categories..</p>
<p>
scenes better than the faces...</p>
<p>
K. suggest maybe change location of PFC stim (eg R PFC more involved with faces)</p>
<p>
Q abt encoding efficacy in relation to phase...
and how does it relate to optimal coding in real world.</p>
<p>
</p>