We have performed group ICA in two groups of patients (BOLD fMRI data). One group is controls, one is with very severe developmental brain abnormalities (mixed group). In controls, ICA revealed 6 components (mostly bilateral), while in the diseased group, we get 30-40 components on the group level, and the components are small and focal. Data preprocessing is the same, data quality is also the same (although the brains themselves might show some anatomical heterogeneity as well).
I am very curious how this excessive number of IC can be interpreted. We may assume that the diseased group have impaired brain functioning, even no brain functioning (e.g. due to neural migration disorders) at the respective areas.
Thank you.
András
There is very little that the number of components found could tell you about brain functioning or connectivity, since many of the components could be related to noise, artefacts or be vascular components. You should take some considerations into account. Did you perform group ICA with temporal concatenation? Did you use automatic estimation of the number of ICs?
If you really want to compare the integrity of the resting state networks in a group of patients vs. controls your approach is not adequate. I suggest that you go for a entire sample group analysis, and after identification of the real resting state networks you apply a two-groups differences analysis. Take into account structural differences into account. In the end, differences found would reflect differences in network integrity/pattern.
Dear Jorge,
Thank you for your answer.
It was concatenated group ICA with melodic, automatic estimation of components. I fully understand that you suggest to use something like "dual regression" on the whole cohort, which is an efficient and general way to analyse such data and test differences in IC-s. This is indeed the practice in many studies we perform.
These subjects are fetuses, with under-developed brain functions. The different cohorts have vastly different anatomy and neural maturation/integrity. Running a whole group concatenated ICA would assume that they have similar networks to a certain extent, hypothesizing only that the strength and spatial extent of these networks are different between the two groups (this is what randomise would test during dual regression). In our case, this assumption do not hold, the two groups may have different RSNs, or RSNs might not exist at all in one of the groups due to missing neurons in the cortex. This is a reproduction of the methodology in http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3868398/.
Still, I am trying to figure out why this specific group shows this phenomenon, e.g. even if these are artifacts and vascular components, why there is a such a big difference compared to controls and how to control this effect in a study.
Best,
András
Have you regressed out motion artifacts before performing group ICA? Excessive and random motion patterns in under-developed fetuses could exist and be quite different from controls. Data cleaning in temporal domain such as using frame displacement to screen out outliers of each time points might be helpful.
This is a case where Seed-based analyses may be your friend to get an idea of what's going on. In particular, I would try simply creating correlation maps from primary motor cortex in each subject and see what you get from visual comparison and group t-test. It will be easier to interrogate data quality and perform focused network analysis that might help clarify the difference in the Melodic ICA recognition...
Dear Andras,
Have you had a look at the run-to-run reproducibility of te ICA estimation?
Here they discuss the problem and propose several techniques to handle it:
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0027594
Best, Markus
Jorge Arrubla I have performed group ICA on my data using Melodic, and reduced it to 30 components. After that, I did dual regression, and filtered the RSNs that I need to work with. I need to see if there are any differences in the two groups. How do I perform this analysis?
- Badges
- Science topic
- Similar topics
- Filing