1 / 56

Independent Component Clustering

Independent Component Clustering. Why cluster components? EEGLAB studysets: the STUDY structure Creating a STUDY – pop_createstudy() Preparing the clustering measures – pop_preclust() Performing the clustering – pop_clust() Reviewing the results – pop_clustedit()

lolitae
Download Presentation

Independent Component Clustering

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Independent Component Clustering • Why cluster components? • EEGLAB studysets: the STUDY structure • Creating a STUDY – pop_createstudy() • Preparing the clustering measures – pop_preclust() • Performing the clustering – pop_clust() • Reviewing the results – pop_clustedit() • Cluster editing – pop_clustedit() • Hierarchical clustering • Signal processing cluster activities – cls_envtopo() • Clustering example – STUDY of 5 datasets EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  2. Why cluster components? • ICA transforms the data from a channel basis (activity recorded at each channel) • to a component basis (activity computed at each independent spatially-filtered cortical or non-cortical component process). • Normally, EEG researchers assume that electrode, say F7 == F7 == F7 ... in each subject – and then ‘cluster‘ their data by channel ... • But this is only roughly correct! EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  3. Example: First Subject Electrode F7 Cortex ∑ IC2 Scalp ICs IC3 IC1 IC4 S. Makeig, 2005 EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  4. Second Subject ∑ Electrode F7 Cortex IC2 Scalp ICs IC3 IC1 IC4 S. Makeig, 2005 EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  5. Third Subject Electrode F7 Cortex ∑ IC2 Scalp ICs IC3 IC1 IC4 S. Makeig, 2005 EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  6. Fourth Subject Electrode F7 Cortex ∑ IC2 Scalp ICs IC3 IC1 IC4 S. Makeig, 2005 EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  7. Largest 30 independent components (single subject) EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  8. So how to cluster components? • The same problems hold for clustering independent components • Across Ss, components don’t even have “the same” scalp maps! •  Are “the same” components found across subjects? • What should define “the same” (i.e., “component equivalence”)? • Similar scalp maps? • Similar cortical or 3-D equivalent dipole locations? • Similar activity power spectra? • Similar ERPs? • Similar ERSPs? • Similar ITCs? • OR …, Similar combinations of the above? … EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  9. Does the spatial distribution of independent components depend on the task the subject performs? i.e. Do “the same” components (and clusters) appear for every task? EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  10. Equivalent dipole density Sternberg letter memory task EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Onton et al., ‘05 Onton et al., 2005

  11. Equivalent dipole density Letter twoback with feedback EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Onton et al., 2005 Onton et al., ‘05

  12. Equivalent dipole density Auditory oddball plus novel sounds EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Onton et al., 2005 Onton et al., ‘05

  13. Equivalent dipole density Emotion imagery task EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Onton et al., ‘05 Onton et al., 2005

  14. Equivalent dipole density Exp I Word memory (old/new) task EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Onton et al., ‘05 Onton et al., 2005

  15. Equivalent dipole density Exp II Visually cued button press task EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Onton et al., ‘05 Onton et al., 2005

  16. Here, The electrode locations were not individualized. MR images were not available  co-registration crude. Single versus dual-dipole model selection was subjective. Different electrode montages  possible location effects … Some caveats EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Onton, 2005

  17. Clustering by spectra (1 subject) Makeig et al., unpublished EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  18. Visual Selective Attention Task 15 subjects EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  19. EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  20. Left mu Right mu Clustering ICA components by eye EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Makeig et al., submitted

  21. Semi-automated clustering • Clustered components from 15 Ss using a ‘component distance metric’ incorporating differences between their (weighted) scalp maps, dipole locations, spectra, ERP, ERSP, and ITC patterns. • Hand-adjusted clusters to remove outliers. • Determined time/frequency regions of significant ERSP and ITC for each component using permutation-based statistics. • Used binomial statistics to highlight time/frequency regions significantly active within clusters. EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  22. 60 50 40 60 50 40 60 50 40 10 10 10 20 20 20 ΘFC LC LC + mean - N1 Component Clusters 9 Ss 11 Comp 13 Ss 23 Comp μLC μRC CM Relative Power (dB) 9 Ss 15 Comp 8 Ss 9 Comp 10 Ss 13 Comp αCP αLP αRP 8 Ss 12 Comp 9 Ss 11 Comp 11 Ss 17 Comp Frequency (Hz) EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering Makeig et al., Science 2002

  23. EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  24. EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  25. EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  26. EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  27. EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  28. Complex event-related dynamics underlie ‘the’ P300 EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  29. Distributed theta burst event - brainmovie3d() EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  30. A FM cluster during working memory Onton et al., NeuroImage 2005 EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  31. The FM cluster accounts for all the memory effect Onton et al., NeuroImage 2005 EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  32. Are obtained component clusters “real“? • Naïve realism (a.k.a. “expertise”) • “Yes! … because I know one when I see one!” • “If it appears where Mu components appear, • and acts like Mu components act, • then it IS a Mu component!” • Convergent evidence (a.k.a., “doublechecking”) • Two possible approaches: • Cluster on PLACE  Check ACTIVITY consistency (re task) • Cluster on ACTIVITY  Check PLACE consistency • Absolute truth: • More ideal forward and inverse models • Invasive multiscale recordings + modeling EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  33. Should all subjects be included in each cluster? • Not all subjects contribute components to each cluster. • Why not? • Different numbers of artifact components (~INR) • Subject differences!? • Is my subject group a Gaussian cloud?? •  subject space EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  34. Known Uncertainty EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  35. The goal of cognitive neuroscience Cognitive Events Phys. Data Re-evaluation Multiple Modes Anticipation Distributed Dynamics Imagination Perturbations Modulations Value Model Individual Subjects Culture Personality Genetics Age / Gender EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  36. Identify a set of datasets as an EEGLAB study or ‘studyset’. Specify the subject group, subject code, condition and session of each dataset in the study. Identify components to cluster in each study dataset. Decide on component measures to use in clustering the study and/or to evaluate the obtained component clusters. Compute the component measures for each study dataset. Cluster the components on these component measures. Review the obtained clusters (e.g., their scalp maps, dipoles, and activity measures). Edit the clusters (manually remove/shift components, make sub-clusters, merge clusters, re-cluster). Perform signal processing within or between selected clusters. EEGLAB clustering procedure EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  37. Cluster on multiple measures (dipole locations, scalp maps, spectra, ERPs, ITCs, ERSPs) in one or more conditions. Reduce the dimension of each measure to a principal component subspace. Compose a PCA-reduced position vector for each component. Cluster the composed component vectors using k-means or other. Use the computed component measures (not PCA-reduced) to visualize the activities and spatial properties of the clustered components. Compute and visualize the cluster-mean measures. Use the clustered study set data as input into std_ functions. Clustering strategy EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  38. Cluster ERP contributions - clust_envtopo() clust_envtopo(STUDY, ALLEEG, 'clusters', [], 'subclus',[3 7 18 20], 'env_erp', 'all', ‘vert', -1100, 'baseline', [-200 0], 'diff', [2 1], 'limits', [-1300 1500 -5 5], 'only_precomp', 'on', 'clustnums' , -4, 'limcontrib', [300 600]); EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  39. Component clustering research issues • Alternative clustering methods • Clustering goodness-of-fit • Cluster plotting details • Add new pre-clustering measures • Study subject differences!? • EEGLAB study design issues • Vector of conditions  more general exp. designs • More complete study statistics EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  40. EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  41. Create a STUDY - pop_study() >> [ STUDY ALLEEG] = pop_study(ALLEEG); EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  42. Compute cluster measures - pop_preclust() >> [ALLEEG, STUDY] = pop_preclust(ALLEEG, STUDY); EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  43. Perform clustering - pop_clust() >> [STUDY] = pop_clust(ALLEEG, STUDY); EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  44. Plot cluster-mean scalp maps - pop_culstedit() >> [STUDY] = pop_clustedit(ALLEEG, STUDY, [2:22]); EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  45. Plot the properties of a mu rhythm cluster EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  46. Rename a cluster EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  47. Merge clusters EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  48. Remove components ( outliers cluster #2) EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  49. Reassign components to clusters EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

  50. Split a cluster - pop_preclust() >> [ALLEEG, STUDY] = pop_preclust(ALLEEG, STUDY, 14 ); EEGLAB Workshop IV, June 26-29, 2007, Aspet, France: Scott Makeig – Component Clustering

More Related