Phy - how to get an overview of manual sorting work done...?

219 views
Skip to first unread message

Tahl Holtzman - Cambridge NeuroTech

unread,
Sep 9, 2015, 6:27:04 PM9/9/15
to phy-...@googlegroups.com
I've just starting playing with Phy and I like what I've seen so far. Maybe I'm doing something wrong, but how can I tell when I've manually sorted through all of the clusters? I.e. in Klustaviewa the Cluster View pane was really helpful for seeing how many clusters ended up in each category, and of course for indicating how much work one still had to invest on a particular file.... and how can I review all of the categorisations I've made, i.e. visualise all clusters I've put into MUA / noise / good...?

On a related note, the ability to plot spike waveforms from many clusters (i.e. >2) was also helpful in gaining a sense of overall file / data quality... ie. something like this:

Inline images 1

As always, I'm excited by the continual development and improvements you guys are working on. With thanks,
Tahl H.


Tahl Holtzman PhD., Founder & CEO
cambridgeneurotech.com


Max Hunter

unread,
Sep 9, 2015, 6:35:53 PM9/9/15
to Tahl Holtzman - Cambridge NeuroTech, phy-users
Hi Tahl,

We're working on something which will show you an overview of the clustering process! You can of course use the command-line, but I agree that something a little bit more functional would be helpful in the GUI.

(for example, launch your session in interactive mode with phy cluster-manual -i myfile.kwik then type model.cluster_groups to list all the clusters and their group numbers - 0 = noise, 1 = MUA, 2 = good, 3 = unsorted)

You can plot more than 2 clusters by manually selecting the ones you want with the use of snippets: for example, to plot clusters 10 and 12, you can type

:c 10 12 [ENTER]

or for 10 through 20, you can do

:c 10-20 [ENTER].

Hope that helps! Thanks very much for your feedback - and that's a great figure :)

Max

--
You received this message because you are subscribed to the Google Groups "phy-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to phy-users+...@googlegroups.com.
To post to this group, send email to phy-...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/phy-users/CAF2U3A1C_DGhuqpVytQPB6S2KC0sCtmC%2BZvQR3k_3iVOrReG%3DA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Tahl Holtzman - Cambridge NeuroTech

unread,
Sep 10, 2015, 5:51:41 AM9/10/15
to Max Hunter, phy-users
Hey Max - thanks for the quick reply - always appreciated!

re: model.cluster_groups - cool, that's helpful. When the GUI overview module is ready I'd be very happy to beta-test and offer feedback.

re: snippets... couldn't get that working... -i throws a syntax error when I type your commands, sat :c 10 12 for plotting snippets from clusters 10 and 12 - any thoughts?!

Cheers,
Tahl


Tahl Holtzman PhD., Founder & CEO
cambridgeneurotech.com



Nick Steinmetz

unread,
Sep 10, 2015, 5:56:41 AM9/10/15
to Tahl Holtzman - Cambridge NeuroTech, Max Hunter, phy-users
What error do you get when you try the snippet? Let me try to be slightly more specific about how to use it: run cluster-manual gui as normal. When you have the main window selected (not the command line), just type 

:c 10 12

then press enter. Of course there needs to be a cluster 10 and 12 in your dataset. Then, to clarify, the result will be to select clusters 10 and 12, just the same way two clusters are "selected" when you use the wizard. (The text command using a colon is what's called a "snippet" - it's not that the function will show you a "snippet" of the data from those clusters).

Does that work? Good luck - 

Nick



Tahl Holtzman - Cambridge NeuroTech

unread,
Sep 10, 2015, 6:06:16 AM9/10/15
to Nick Steinmetz, Max Hunter, phy-users
Working nicely now - thanks for your help Nick.

I look forward to seeing some of your 100's of channels probe data pushed through Phy in due course.

Best,
Tahl

Tahl Holtzman PhD., Founder & CEO
cambridgeneurotech.com



Reply all
Reply to author
Forward
0 new messages