BEP016: Diffusion derivatives

7 Members
2 Servers

Load older messages

Timestamp Message
23 Aug 2018
18:38:54@jchoude:matrix.orgjchoudeok thanks
18:43:11@jchoude:matrix.orgjchoude invited @michiel_cottaar:matrix.orgmichiel_cottaar.
18:43:41@michiel_cottaar:matrix.orgmichiel_cottaar joined the room.
18:44:19@michiel_cottaar:matrix.orgmichiel_cottaarHi, thanks for inviting me
18:44:35@jchoude:matrix.orgjchoudeHi, nice to have you in the discussion
18:44:48@arokem:matrix.orgarokem joined the room.
18:45:06@arokem:matrix.orgarokemI am in the Diffusion derivatives room
18:45:16@jchoude:matrix.orgjchoudeCurrently, the discussion is still in the main BEP group, about file naming and suffixes. When we get to proper DWI-related discussions we'll update you
18:45:28@arokem:matrix.orgarokemRedacted or Malformed Event
18:45:40@michiel_cottaar:matrix.orgmichiel_cottaarThanks, let me know
18:46:26@jchoude:matrix.orgjchoudeDo you have the link for the current (unfinished) draft of the DWI spec?
18:47:22@jchoude:matrix.orgjchoudeIf you want to have a look in the meantime. We mainly focused on the "models" section up to now, trying to list the mainstream models
18:47:53@jchoude:matrix.orgjchoudeThe goal is not necessarily to identify every single model ever published, but more what is common in this first phase
19:21:12@michiel_cottaar:matrix.orgmichiel_cottaarIt's up to you guys, but I'm a bit worried that focusing on the "models" section is a bit of a rabbit hole. If I am developing a tool relying on the output of diffusion MRI models, I would mostly want standard locations for model-derived maps (for statistical analysis), for ODFs (dyads or spherical harmonics for tractography), and/or tractography streamlines (for connectivity analysis/ROIs), which being described in the other sections. I can not imagine there will be a lot of tools that will work on the diffmodels directly rather than the maps or ODFs. So in my mind diffmodels is actually an intermediate file format and I'm not sure if we really need to define the diffmodels for the initial release of the DWI spec.
19:54:40@jchoude:matrix.orgjchoudeYou're right that the diffmodels are an intermediate step. We worked on them since the section was already there, and I think it's still important to have some support for it since they can be useful derivatives (say, if someone wants to extract new metrics from a precomputed model).
19:55:19@jchoude:matrix.orgjchoudeAlso, our main focus this afternoon will be on the streamlines formats, and tomorrow will be the rest of derived maps (ODFs, metrics)
20:18:13@jchoude:matrix.orgjchoude Ariel, Franco, where are you? Can't seem to find you
20:48:07@michiel_cottaar:matrix.orgmichiel_cottaarI'll probably call it a night. I don't have much experience with streamline formats anyway. I won't be able to come online tomorrow during your afternoon, but I'll keep an eye on the google docs. It looks like there is good progress. Good luck in keeping the momentum going and enjoy yourselves.
20:48:26@jchoude:matrix.orgjchoudeYes, thanks for the feedback and your time!
21:32:52@jchoude:matrix.orgjchoudedest_dtype = np.dtype([('R', 'uint8'), ('G', 'uint8'), ('B', 'uint8')]) out_data = np.zeros(original_dat.shape[:3], dtype=dest_dtype)
for id in np.ndindex(original_dat.shape[:3]):
    val = original_dat[id]
    out_data[id] = (val[0], val[1], val[2])
21:33:34@jchoude:matrix.orgjchoudenew_hdr = original_im.get_header() new_hdr['dim'][4] = 1 new_hdr.set_intent(1001, name='Color FA') new_hdr.set_data_dtype(dest_dtype)
24 Aug 2018
16:26:02@franco-p:matrix.orgfranco-pestilli jchoude: some edits added to proposal
22:58:57@filo:matrix.orgChris Gorgolewskiheads up - we added 'SkullStripped' as a mandatory field for all preprocessed files
23:03:12@jchoude:matrix.orgjchoudeSince this is in the common derivatives document, do you still want us to specify it, or it's gonna be explicit when everything is merged

There are no newer messages yet.

Back to Room List