Hi Alfonso 
First of all thank you to the team for making such a wonderful toolbox. 
I have few queries regarding denoising and frequency decomposition. 

I ran Conn analysis to check for any frequency dependent effects on  a particular patient group ( with filter bank n =4, which created four  additional frequency bands in the Condition set up. After analyzing the results, I could see there are  newer and useful fmri activations  seen  in each of the four conditions created than that were  seen in  the original frequency band ( default frequency 0.008Hz - 0.09Hz). I also ran analysis with filter bank n = 3 and 5. As the conditions increases, I have more number of unique activations which could not be seen when the original rest condition frequency alone was applied. My doubts are : 


1. Can this technique be  seen as an advantage and reported as better  method compared to  the original full band frequency ( i.e , no decomposition)?
2. Could the frequency decomposition technique with   discrete cosine transformation used in conn be attributed to this?. Is there any article which explains the use of sub band coding using DCT in conn which  could be cited?

3. Could Frequency decomposition into equal bands by dct be interpreted as a better approach in our study ? Or could this be some noise / artifacts or other phenomena we are not aware of?

Your valuable suggestion and advice is highly appreciated. 
Thanks 
Jos