Lecture 2 Question: Drug Therapies

[ajulian3]

Would it be possible from a graphical perspective to map drug treatments across different subjects using time as one of the independent variables along with neuron frequency as the dependent variables and then create a statistical test to match against treatments? If so, which algorithms and statistical analytics platforms would be best to accomplish this task?

[mrjiaruiwang]

Well we could try to convert different timeframes into maps. We can design a function that takes neuron frequency data and convert it into a graph, like we talked about in class. So for example, you could define an edge if one neuron activates near the same time as another. Then analyze the graph by some algorithm, like clustering, and compare it across different timescales. For example, you could plot the absolute change from clustering across the timescale to see if you can find any patterns or similarities between the drug concentration.

[mblohr]

Perhaps a brain-graph state space model, such as implemented by Sarma et al for epileptic brain states, but applied to brain states undergoing drug treatment, would be useful.