clustering of ica components - sccn.ucsd.edu
TRANSCRIPT
![Page 1: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/1.jpg)
Clustering of ICA components
Arnaud Delorme
(with Julie Onton, Romain Grandchamp, Nima Bigdely Shamlo, Scott Makeig)
![Page 2: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/2.jpg)
Steps of clustering
• Select ICA components for clustering
• Precompute measures of interest
• Cluster measures
• Plot clusters and edit them if necessary
![Page 3: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/3.jpg)
Edit dataset info
![Page 4: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/4.jpg)
ICs to cluster
![Page 5: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/5.jpg)
Precompute data measures
![Page 6: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/6.jpg)
Precompute data measures
TIP: Compute all measures so you can
test different combinations for clustering
Time-frequency options
![Page 7: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/7.jpg)
Cluster components
![Page 8: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/8.jpg)
ICs (all subj)
ERSP
(t
ime/
freq
)
PC weights * PC
tem
plat
es
PCA Mean ERSPs
Precluster schematic
ICs (all subj)
Pow
er
spec
trum
PC weights * PC
tem
plat
es
PCA Mean spectra
3D dipole position
# PCs
ICs
(all
subj
)
# PCs concatenate IC
s (a
ll su
bj)
PC t
empl
ates
x3
ICs
(all
subj
)
PC weights *
concatenate
ERSP
Spec
trum
Dip
oles
ICs
(all
subj
)
PCA
ICs
(all
subj
)
![Page 9: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/9.jpg)
Precluster: Use singular values from PCA
10% of max singular value
~ relative variance of
principal components
ICs (all subj)
ERSP
(t
ime/
freq
)
PC
tem
plat
es
PCA Mean ERSPs
ERSP
(t
ime/
freq
)
# PCs
![Page 10: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/10.jpg)
Precluster schematic
ERSP
Spec
trum
Dip
oles
ICs
(all
subj
)
ICs
(all
subj
)
OR
Each component is a dot Clustering will group these dots
![Page 11: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/11.jpg)
1. k initial "means" (in this case k=3, (shown in color)) are randomly selected from the data set (shown in grey).
Classical KMean
2. k clusters are created by associating every observation with the nearest mean.
3. The centroid of each of the k clusters becomes the new means
4. Steps 2 and 3 are repeated until convergence has been reached.
![Page 12: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/12.jpg)
Cluster components
![Page 13: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/13.jpg)
What measure(s) should you use?
It depends on your final cluster criteria… - If for example, your priority is dipole location, then cluster only based on dipole location…
But consider:
- What is the difference between these two components?
Choosing data measures
![Page 14: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/14.jpg)
Choosing data measures
Similar dipole location, very different orientation.
Obvious dramatic effect on scalp map topography:
But, do they perform the same functions?
![Page 15: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/15.jpg)
Choosing data measures
ERPs seem different…
![Page 16: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/16.jpg)
Subject differences?
![Page 17: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/17.jpg)
Subject differences?
![Page 18: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/18.jpg)
Subject differences?
![Page 19: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/19.jpg)
Plot/edit clusters
![Page 20: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/20.jpg)
Plot cluster data
Plot mean scalp maps for easy reference
![Page 21: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/21.jpg)
Plot cluster data
Choose which
cluster
Choose which components
![Page 22: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/22.jpg)
Plot cluster data
![Page 23: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/23.jpg)
Plot cluster ERP
![Page 24: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/24.jpg)
Reassigning components
![Page 25: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/25.jpg)
Issue with standard clustering
![Page 26: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/26.jpg)
Measure projection
(EEGLAB extension by Nima Bigdely Shamlo) only has one pre-clustering parameter.
(Affinity clustering by Pernet, Martinez, Delorme)
![Page 27: Clustering of ICA components - sccn.ucsd.edu](https://reader033.vdocument.in/reader033/viewer/2022052322/628a52365a291144b214cf4d/html5/thumbnails/27.jpg)
Exercise
• Load the STUDY • Precompute ERP, power spectrum and scalp
topographies • Precluster and cluster components using spectrum and
dipoles location • Look at your cluster. Identify frontal midline theta cluster
and occipital alpha cluster • Plot significant difference for one component cluster
spectrum between the two conditions in the default design