LIDNET Alpha Asymmetries
Average cross sections of Alpha power at single
electrodes for eyes open (o) and eyes closed (x). Note that there are only two
samples for the Spline CSD reference. Approximate values for 32 channel
electrocap data (EO/EC means) from Hagemann's preprint are indicated by plus
signs (+) for Cz and Ave Ref.
Average cross sections of Alpha power at single
electrodes for the different references before(o)and after BESA EOG correction(x).
Note that there are only two samples for the Spline CSD referenced data before
correction and thirty-six after BESA EOG correction. Approximate values for 32
channel electrocap data (EO/EC means) from Hagemann's preprint are indicated by
plus signs (+) for Cz and Ave Ref.
Where's the Alpha power? Hagemann shows for 32 channel electrocap data almost a ten fold increase in Alpha power from central to occipital areas whereas the net data Alpha power profile is nearly flat between these areas. One possibility I've been considering is that the electrodes themselves increase the scalp conductivity even without any electrolyte bridging. The 128 electrodes cover approximately 13% of the area with a conductivity at least an order of magnitude greater than the scalp's (0.33 mho/m compared to 3.3 for a 2% NaCl solution). Based on a simple area estimate the effective scalp conductivity would become 3.3*0.13 + 0.33*0.87 = 0.72 mho/m thus more than doubling. Next running a simple dipole model in Patrick Berg's SimDip program shows that doubling the scalp's conductivity approximately halves the relative power difference between central and occipital areas.
Jurgen Kayser (personal communication) reports dramatically smaller EOG signals with the GSN system compared to an ElectroCap setup on the same individuals. Again using the SimDip program with blink dipoles shows about a third decrease in the measured EOG signal with a doubling of the scalp conductivity. In a study designed to compare GSN 128 channel with conventional 30 channel ERP's Kayser et al. (2000) report an agreement between latencies (N1, N2, P3) for both group and individual averages. "However, all components were reduced for the 128-channel system, with the extent of the reduction varying considerably across electrodes, conditions, and subjects." Also, the 128-channel data failed to reveal one of the hemispheric asymmetries brought out in the 30-channel data.
Although this theory might explain why the Alpha power is more subdued in the net data, it doesn't provide any clear direction into how to better process the net data in order to get good asymmetry estimates. However, it would be important to quantify this effect if it is real in order to get better modeling results from LORETA and BESA. It will be interesting to see if the Alpha power is even more subdued in the 256 channel net data.
TEST-RETEST CORRELATIONS
SINGLE ELECTRODES
Original data (No Mask)
N=175 (from Chris) N<=18
CZ AVE CZ
AVE SPLAVE
FP1L .14 .16
-.32(-.33) .09 (.08) .26 (.11)
F3L .18 .22
.09 (.18) -.04(-.03) .06 (.03)
F7L .16 .13
-.02 (.01) -.38(-.35) -.42(-.34)
T3L .18 .09
.05 (.03) .19 (.16) .11 (.02)
C3L .25 .25
.35 (.34) .31 (.29) .34 (.32)
T5L .37 .12
.46 (.44) -.13(-.19) -.09(-.20)
P3L .47 .32
.29 (.29) .07 (.07) .16 (.10)
O1L .34 .15
.22 (.19) -.28(-.27) -.26(-.23)
SINGLE ELECTRODES BESA eye corrected
data (No Mask)
N=175 (from Chris) N<=18
CZ AVE CZ
AVE SPLAVE
SPLCSD
FP1L .14 .16
-.28(-.39) -.06(-.14) -.01(-.13) -.13
F3L .18 .22
-.13(-.09) -.05 (.01) .15 (.17) .01
F7L .16 .13
-.26(-.42) -.16(-.34) -.12(-.37) -.10
T3L .18 .09
-.04(-.08) .06 (.01) .09 (.02)
.14
C3L .25 .25
.41 (.39) .45 (.48*) .50*(.51*) .40
T5L .37 .12
.47 (.44) .13(-.05) .21 (.05)
.57*
P3L .47 .32
.40 (.42) .31 (.38) .37 (.41)
.41
O1L .34 .15
.31 (.31) .04 (.05) .03 (.07) -.04
* significant at 5% level
Plots below show correlations for masked data
TEST-RETEST CORRELATIONS
GROUP ELECTRODES
Original data (No Mask)
N=175 (from Chris) N=18
CZ AVE
CZ AVE
SPLAVE
FP1L .27 .31
.46 (.52*) .62*(.56*) .59*(.46)
F3L .24 .25
.21 (.24) -.13(-.10) -.15(-.15)
F7L .21 .31
.47*(.48*) .12 (.13) -.20(-.02)
T3L .28 .12
.12 (.17) .37 (.32) .41 (.31)
C3L .32 .24
.35 (.35) .32 (.30) .43 (.37)
T5L .44 .18
.32 (.29) -.13(-.20) .00(-.12)
P3L .52 .34
.37 (.35) .02(-.03) .08 (.02)
O1L .42 .24
.17 (.14) -.25(-.33) -.22(-.28)
GROUP ELECTRODES BESA eye
corrected data (No Mask)
N=175 (from Chris) N=18
CZ AVE CZ
AVE SPLAVE
SPLCSD
FP1L .27 .31
-.06 (.03) .35 (.38) .45 (.47*) .12
F3L .24 .25
.17 (.16) -.01 (.06) .22 (.27) .29
F7L .21 .31
-.30(-.23) -.21(-.16) -.14(-.08) -.17
T3L .28 .12
.28 (.34) .60*(.67*) .54*(.60*) .11
C3L .32 .24
.40 (.38) .50*(.52*) .62*(.64*) .21
T5L .44 .18
.33 (.30) .12 (.01) .14 (.08)
.49*
P3L .52 .34
.42 (.41) .34 (.28) .35 (.34)
.07
O1L .42 .24
.21 (.20) .13 (.00) .14 (.04)
.07
* significant at 5% level
Correlation of asymmetry measures between different
reference montages:
Hagemann (32 electrodes)
FP1 F7L F3L
T3L C3L T5L
P3L O1L
CZ x AVE
.36 .63 .35
.77 -.11 .99
.96 .99
LIDNET original (N=36)
Group Electrodes
FP1 F7L F3L
T3L C3L T5L
P3L O1L
CZ x AVEALL
.83 .70 .72
.58 .36 .41
.20 .08
x SPLAVE
.78 .59 .51
.27 .27 .31
.18 .03
AVEALL
x SPLAVE
.99 .92 .87
.85 .97 .98
.99 .99
LIDNET BESA EOGC (N=36)
Group Electrodes
FP1 F7L F3L
T3L C3L T5L
P3L O1L
CZ x AVEALL
.68 .90 .60
.50 .17 .32
.18 .14
x SPLAVE
.55 .77 .24
.18 .10 .22
.28 .10
AVEALL
x SPLAVE
.91 .93 .82
.79 .94 .97
.99 .98
Correlations between Original LIDNET and BESA EOGC
LIDNET (N=36)
Group Electrodes
FP1 F7L F3L
T3L C3L T5L
P3L O1L
CZ
.22 .28 .80
.87 .95 .96
.98 .98
AVEALL .53 .42 .78 .81 .86 .85 .92 .92
SPLAVE .63 .37 .69 .87 .86 .85 .93 .93
Correlations between Masked and not Masked Original
LIDNET and BESA EOGC LIDNET (N=36)
Group Electrodes
FP1 F7L F3L
T3L C3L T5L
P3L O1L
ORIG CZ
.95 .99 .98
.98 .99 .97
.99 .99
BESA EOGC CZ
.97 .98 .98
.97 .99 .98
.99 .99
ORIG AVEALL
.95 .99 .97
.96 .99 .96
.99 .99
BESA EOGC AVEALL .98
.99 .98 .95
.99 .94 .99
.99
ORIG SPLAVE
.94 .96 .95
.93 .98 .96
.99 .99
BESA EOGC SPLAVE .96
.97 .97 .92
.99 .95 .99
.99
Original LIDNET data
Single electrodes (rows from top to bottom :FP1L, F3L, F7L, T3L, C3L, T5L, P3L,
O1L)
CZ
AVE
SPLAVE
BESA EOG corrected LIDNET data
Single electrodes (rows from top to bottom :FP1L, F3L, F7L, T3L, C3L, T5L, P3L,
O1L)
CZ
AVE
SPLAVE
SPLCSD
Original LIDNET data
Group electrodes (rows from top to bottom :FP1L, F3L, F7L, T3L, C3L, T5L, P3L,
O1L)
CZ
AVE
SPLAVE
BESA
EOG corrected LIDNET data
Group electrodes (rows from top to bottom :FP1L, F3L, F7L, T3L, C3L, T5L, P3L,
O1L)
CZ
AVE
SPLAVE
SPLCSD
Recommended data path:
1) Eye correct with BESA
2) Score out remaining artifacts
3) Compute PSD with respect to lots of references (CZ, MAST?,
AVEALL, SPLAVE, SPLCSD which looks promising for T5L)