jovo
commented
9 years ago very cool! thanks for the clarification!
On Thursday, April 30, 2015, SandyaS72 notifications@github.com wrote:
So in case anyone is interested in what's happening with that project now, Austin and I have both continued with working on it. In fact, Sam's work was kind of our precursor. So basically, what Sam saw is that when he found these discrete states during seizure, there were specific states (I don't remember which numbers from his diagram offhand) that were "foci-hot" or "foci-cold" in patients with successful resections, meaning that the foci (assumed to be correct because the person stopped having seizures afterwards, or at least whatever they resected contained the foci) had either very high centrality or very low centrality in the network in those specific states. This basically led to the hypothesis that during seizure, the ictal onset zone might have a very specific signature or evolution in its centrality that's distinguishable from the other nodes. What we defined as the centrality "signature" of each electrode was the following: Get the eigenvector centrality vector for each time window as described in the presentation today, then convert the EVC vectors to ranked vectors (from 1 to n - because we only cared about the relative centrality of the nodes). For each electrode you now have a ranked centrality signature through time. Results: What we saw is that there was a pretty consistent pattern in the centrality signatures of the ictal onset zone (based on patients with successful resections again). At the onset of seizure, their centralities would suddenly drop to being very not central in the network and then later, mid to late seizure, they would come back to be very highly central (which might agree with the idea of that region doing something very different from everything around it at first and then entraining its neighbors during the course of seizure). This was really consistent for seizures within the same patient, and across patients as well. This is super summarized, but this was the next step made before we went forward with trying to automate the algorithm further to make an actual tool with Coulter Foundation funding.
— Reply to this email directly or view it on GitHub https://github.com/openconnectome/Statistical-Connectomics-Class/issues/254 .
the glass is all full: half water, half air. openconnecto.me, jovo.me, office hours https://www.google.com/calendar/embed?src=e2ktu4lrgul8anp8hclrcminp8%40group.calendar.google.com&ctz=America/New_York
maxcollard
mblohr
akim1
So in case anyone is interested in what's happening with that project now, Austin and I have both continued with working on it. In fact, Sam's work was kind of our precursor. So basically, what Sam saw is that when he found these discrete states during seizure, there were specific states (I don't remember which numbers from his diagram offhand) that were "foci-hot" or "foci-cold" in patients with successful resections, meaning that the foci (assumed to be correct because the person stopped having seizures afterwards, or at least whatever they resected contained the foci) had either very high centrality or very low centrality in the network in those specific states. This basically led to the hypothesis that during seizure, the ictal onset zone might have a very specific signature or evolution in its centrality that's distinguishable from the other nodes. What we defined as the centrality "signature" of each electrode was the following: Get the eigenvector centrality vector for each time window as described in the presentation today, then convert the EVC vectors to ranked vectors (from 1 to n - because we only cared about the relative centrality of the nodes). For each electrode you now have a ranked centrality signature through time. Results: What we saw is that there was a pretty consistent pattern in the centrality signatures of the ictal onset zone (based on patients with successful resections again). At the onset of seizure, their centralities would suddenly drop to being very not central in the network and then later, mid to late seizure, they would come back to be very highly central (which might agree with the idea of that region doing something very different from everything around it at first and then entraining its neighbors during the course of seizure). This was really consistent for seizures within the same patient, and across patients as well. This is super summarized, but this was the next step made before we went forward with trying to automate the algorithm further to make an actual tool with Coulter Foundation funding.