-
Notifications
You must be signed in to change notification settings - Fork 2
/
notes.txt
152 lines (96 loc) · 4.33 KB
/
notes.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
https://mne.tools/stable/install/advanced.html
file:///home/prakhar/Downloads/fnsys-09-00175.pdf
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9204388/
https://www.sciencedirect.com/science/article/pii/B978012407908300011X
Meaningfulness of connections
connectivity edges is a function of pearson correlation of disparity of dpli across nodes
make sure thresholding works - robustness in dynamic thresholding
global and node level metrics - big picture and small picture
Healthy controls vs taichi - differences, why and how quantified
Analyze the backbone difference and how it affects information flow
granger causality
minimum spanning trees
Processing participants:
- how to pre-process a channel
TODO:
- resilience attack
- icalabel training
- https://labeling.ucsd.edu/tutorial/practice
- meet with max to learn about labelling
- 2 step thresholding
- think about alternative end results
-
Bad channel identification
- scale the chart to see noisy channels. Crzy amplitude variaition
- Drift over time
- Crazy spikes in the Os because of the movement
- Mark bad channel
- run ICA after it (ICA assumes channels are independent, interpolation makes channels dependent)
-
NEW MEETING:
- Subject 116 -
: Rmastoid data replaced with GND.
: Interpolation comes after the ICA
: Get more data?
: We want TCOA against TCOA
: For 116 copy GND to Rmast
- Subhject 117 EO - only 10 components
- Subject 116 EC - 10 comps, EO-12 comps
- autocorrelation + std dev to figure out bad channels
2 step filter
- upper bound is fixed
- disparity filter is after bootstrapping
- bootstrapping is a linear combination of 60 windows
- do bootstrapping per person
- calcultate number and weight of edges across windows and calc difference
- use the diff values (vary a lot), retain ones with
- all written
- Use the bootstrapped copies to get the threshold to look under
- use disparity filter on windows
- averaging after filtering
- Don't do disparity filter for abstract, use jonathan's new code
- set a random threshold, graphs sparsity, keep top 10% connections only
-Subject 202 treated same as 116
12/6/23
Action Items
- Check what's slowing down the computations
- HAL, apply delta over break
- Dr.H wants submit to conf in february, EMBC
- Automatic Bad Channel detection using autocorrelation
- Clean up HAL workspace
- Thresholding operations
- Bootstrapping > disparity >
- Input : Graph, Output: thresholded graph
- Make it builtin
MEETING 12/19
- Work on barebones
- Rerun the 2 participants with 50 components (117 EO and 116 EO)
- Bad channel automation not a priority
- Redo steps after double checking ICA
- Global vs local results is a paper of its own, Graph theoretical analysis is it's own thing
- Split the Paper?
- Start with Functional Connectivity, then Graph Theoretical
Question about functional connectivity analysis:
- dPLI gives directional information
- problem with using just dPLI, is that with resting state data, we don't get much room
to interpret reversal of directional talking.
- questions about general organizaiton of the brain, how do nodes talk to each other? Not
what we are doing.
- We need to answer a question about tai-chi practitioners? The change in directionality
cannot be interpreted in the context of data we have. Reverse Inference flaw in resting state data.
- Characterization and understanding relationships. Dynamical nature of EEG to look at funct connectivity,
fMRI looks at changes in blood flow in the brain, EEG much more 'true' to activity.
Explore amplitude coupling. How there patterns of co-activating. Characterize the
nature of these relationship. Main result is huge changes in node behaviour in different modes
Dipole's Hilbert transform of the amplitude of the envelope of coupling.
Weighted PLI over dPLI: how strongly node i is leading/lagging node j.
What is the strength of coupling in phases vs amplitude? Bands and binding
Coupling in amplitude vs phase changes in resting and active states.
- Change it so that using one person's edges across windows, do that for
all subject's edges across all windows
4XX: older adults TC practitioners; 1XX: young adults; 2XX: older adults non-TC practitioners
Meeting 01/04
- Phase coupling
- Pearson correlation between the envelope of amplitudes
- Check how final values/result change across different thresholds used
- :w