Hi @Yacine Mahdid. I am pulling the Adaptive Reconfiguration manuscript together for our submission to Science Translational Medicine (our aim is to submit by the end of August). I had a discussion with Stefanie today regarding final steps. We have a few last changes still needed in order to take this paper across the finish line. 1) Re-run hubs using only the node degree with the custom threshold (as this better captures the reconfiguration that is able to separate patients who eventually recover from those who do not). 2) We had discussed eliminating the cosine similarity from the Hub-DRI, and computing contrast vectors. This replicates the approach we use with dPLI and eliminates an intermediate step. 3) For the final iteration of the graph that illustrates the separability of the groups (recovered vs. non-recovered) (SEE ATTACHED), we would like to combine the new version of the Hub-DRI (i.e. with hub degree and contrast vectors, instead of cosine similarity) and Attempt #4 of dPLI-DRI, which includes the full dPLI matrices (rather than only fronto-parietal connections) and no weights.
Hi @Yacine Mahdid. I am pulling the Adaptive Reconfiguration manuscript together for our submission to Science Translational Medicine (our aim is to submit by the end of August). I had a discussion with Stefanie today regarding final steps. We have a few last changes still needed in order to take this paper across the finish line. 1) Re-run hubs using only the node degree with the custom threshold (as this better captures the reconfiguration that is able to separate patients who eventually recover from those who do not). 2) We had discussed eliminating the cosine similarity from the Hub-DRI, and computing contrast vectors. This replicates the approach we use with dPLI and eliminates an intermediate step. 3) For the final iteration of the graph that illustrates the separability of the groups (recovered vs. non-recovered) (SEE ATTACHED), we would like to combine the new version of the Hub-DRI (i.e. with hub degree and contrast vectors, instead of cosine similarity) and Attempt #4 of dPLI-DRI, which includes the full dPLI matrices (rather than only fronto-parietal connections) and no weights.