I think you can perform a ROI analysis related in resting state functional network. After, extracting resting state signals, you can perform a power spectrum analysis and determining the most relevant frequency band to your hypothesis. From power spectrum, you can extract the average power in a frequency range or maximum power value and correlate it with behavioral results.

Or a possible other option; you can directly perform a FC analysis and you can correlate the behavioral measure with the correlation coefficient that shows the strength of connection between two regions.

Hi Aykut, thank you for your precious response. Before running the functional connectivity, I was wondering if it should be possible to extract my independent components as numerical vectors and simply proceeding with correlation analysis.

Thank you again,

Edoardo

Hi Edoardo, as far as I understand you are asking that if you can associate with independent components (means time series) and one number (such as age, depression score etc.). You have to extract one value from the independent components.

May be if you explain your problem in details I can help you.

Regards

Aykut

Hi Aykut, I'm sorry but I'm quite naif about this topic. Based on your experience, is it possible to extract a sort of mean of the time series for every component? Anyway, I think that functional connectivity is definitely the best solution for my purposes.

Thank you,

Edo

-is it possible to extract a sort of mean of the time series for every component?

-It is. However, instead of a mean value of a time series, you should use all time series as a input for functional connectivity metric. Mean is not a good option.

Thank you very much, Aykut :)

Hi, if I understand your question well, you can correlate behavioural scores with FEATURES derived from the ICs. These features (alone or in combination) rely on your hypothesis, inquiry et cetera....This allows you to conduct across-subject correlations. Should you need further details please contact me in pvt.

The MANCOVAN option under toolboxes allows you to select your components of interest and correlate with outside things, like age, iq, symptoms, gender, group etc. This can be done with spatial/within-network features, or FNC, or spectra. You can select all ICs or only the ones you're interested in.

http://mialab.mrn.org/software/mancovan/

http://mialab.mrn.org/software/gift/documentation.html

Thank you for all your suggestions :)

Hi,

ICA is a nice method because the data analyst does not have to interfere with the analysis process, e.g. previously assuming a model (which could introduce severe bias, since the beginning). Therefore, it seems a bit contradictory to use ICA and then trying to correlate with whatever it is.

What you may do, in order to preserve the model-free strategy, is to compare two ICA sets acquired in two different conditions (e.g. with therapeutic vs. placebo, or male vs. female, or old vs. young). As each IC has its own timecourse, you just have to compare it with all the ICs’ timecourse of the other set, using any suitable statistic. Of course, if you start to have more conditions, the problem becomes untreatable very easily.

Another method is to use machine learning (e.g. artificial neural networks) for classification purposes. You just have to split your targets into categories. But in this case, for the training stage, you must know which ICs pattern correspond to each category. After building the model, you may start to make predictions (guessing targets) in order to assess if the model is acceptable (or not). You may find here some work that I did with this strategy:

Conference Paper 'Mind Reading': Hitting Cognition by Using ANNs to Analyze f...

I also used ICA with GLM here

Article Investigating the Role of the Ventromedial Prefrontal Cortex...

You use the ICs’ timecourses as independent variables, but you should model the dependent variable with some sort of stimulation. I’m afraid that resting state isn’t a suitable paradigm for this.

Hope this helps. Regards.

Hi Edoardo!

Which kind of analysis do you want to do? Do you have significant group differences in some networks? or do you want to use the network timeseries ?