FAQ

1.Try fast oopsi first 2. Tweak parameters! The most important parameters to tweak to get this working nicely are gamma and lambda. Gamma relates to the decay constant of the calcium transients, and lambda is a measure of the expected firing rate (per frame).

These can be a little tricky to estimate. Lambda is the one to crack first, as it kind of results in a gain control effect. Too high a lambda results in an inferred spike train that just looks like the original fluorescence trace. Too low will result in no spikes being detected. So play with lambda by decreasing it by an order of magnitude at a time (line 129 in the current version of fast_oopsi) until the inferred train goes flat (no spikes); then gradually turn it back up. Here I found that 0.09 worked nicely for my data

if ~isfield(P,'lam'), P.lam = 0.09*ones(V.Ncells,1); end

3. Also, try feeding in raw fluorescence rather than df/F. Lastly, more minor tweaks of gamma may help.