On 10/30/2009 12:12 PM, Danny Kim wrote: > Hi Donna, > > Thanks again for your comments and answers. > When can we expect to see the baby brain atlas available on SumsDB?
A rational person would not be afraid of jinxing a paper's acceptance. As it happens, I'm not all that rational, so afraid I am. It is not inaccurate to say that I hope SumsDB will have a populated directory for this paper by Christmas. That said, we didn't really use a *volumetric* baby atlas for the papers we coathored with Jason Hill, Jeff Neil, Terrie Inder, et. al. I believe Chris Smyser (http://neuro.wustl.edu/aboutus/facultybiographies/smyser.htm), also working with Neil/Inder, Jason Hill and Avi Snyder, may have developed one for his functional connectivity studies. I do not know whether this atlas was or will be made public. > > Appreciate all the help, and I look forward to seeing your group's > work on neonatal atlases! The surface atlases certainly will be made public -- hopefully sooner rather than later. > > Best of wishes and happy Halloween! Mine was good; hope yours was, too. > > > > Danny Kim > Brainmapping > Dept. of Neurology > BC Children's Hospital > Vancouver, BC > > > On Fri, Oct 30, 2009 at 7:12 AM, Donna Dierker > <[email protected] <mailto:[email protected]>> wrote: > > Hi Danny, > > See inline replies below. > > Donna > > > Hi Donna, > > > > thanks for getting back to me quickly :) > > > > I forgot to mention. I created the surface from a T1 image of a > preterm > > baby > > brain. I managed to segment the white matter via utilizing > Freesurfer to > > get > > the bias-corrected T1 normalization image and Brainvisa's anatomic > > pipeline > > to acquire wm/gm masks. Then I used SPM to extract the WM mask > from the > > gm/wm mask. When the mask came out it resembled a paintball > blast mark and > > had the ventricles and the thalami/basal ganglia structures not > > highlighted > > so I did some very quick manual correction. > > > > and thus the freakish surface was born :) > > Okay. Someday Freesurfer may segment baby brains, but currently the > contrasts properties of the baby brain make this a suboptimal > solution. > > > > THANKS for letting me know about the surface being registered to the > > 711-2B > > space and about the ROI volumes. > > Even we haven't tried auto-landmarks on baby brains, and we have no > immediate plans to do so. Nor do we register baby brains to > 711-2B. We > do have some baby atlases we use. > > > > > I'll see if I can register the images to 711-2B space to see if > there is a > > difference. Also I'll try to spend more time in manually filling > in the > > ventricle and correcting the segmentation. > > This segmentation seems beyond manual patching to me. > > > > btw, I saw from the Caret's wiki page that the lab is > collaborating with > > Terrie Inder's group in studying Human cortical development. > > How did you acquire that developing cortical surface image on > the wiki > > page > > (see attached), b/c that looks really good. > > Jason Hill (MSTP student who has completed PhD work and is now in > clinical > rotations) and Andy Knutsen (preparing to defend his thesis) > developed an > algorithm called LIGASE to segment the baby brains. They still needed > some manual patching, but the quality was pretty good, if not > excellent. > I'm pretty sure they used the T2, rather than the T1. > > We are in the process of addressing reviewer comments on a paper > describing these methods and the term control atlas Jason > produced. We > hope to have a citation for you soon. > > Maybe David will post his SfN poster on our ftp site. It covers > this work. > > If you are serious about generating surfaces of baby brains, getting > LIGASE might be worth your while. It runs on matlab. I think > there may > be some documentation. Jason is super busy and not as available as he > might wish to answer user questions, but I can promise the resulting > segmentations and surfaces are nearly as pretty as the city where you > live. > > > > > > > > Cheers, > > > > > > Danny Kim > > Brainmapping > > Dept. of Neurology > > BC Children's Hospital > > Vancouver, BC > > > > > > On Thu, Oct 29, 2009 at 1:45 PM, Donna Dierker > > <[email protected] <mailto:[email protected]>>wrote: > > > >> On 10/29/2009 02:39 PM, Danny Kim wrote: > >> > Hi fellow caret users, > >> > > >> > This is my first time posting a question to the mailing list. > Forgive > >> > me if I leave out anything important or formality. I'll try > to include > >> > everything I know. > >> > > >> > so I've been trying to landmark and paint the sulci of this > brain (see > >> > attached photos of Lateral and Medial view of resulted > fiducial and > >> > inflated surfaces). > >> Well, it's almost Halloween, so I guess scary images are seasonally > >> appropriate. But I must say, that is a pretty hideous fiducial > surface > >> you have there -- lots of appendages and handle-like spindles. > Did you > >> tessellate something like a FSL fast segmentation? > >> > > >> > I keep getting paint Sul.CeS not found and the automatic landmark > >> > labeling does not go to finish in surefit segmentation. > >> I just got that error today myself. In my case, it was because > I was > >> running auto-landmarks and feeding it a bad Depth > column-name/number. > >> But you might also see that error if your spec file has more > than one > >> anatomical volume in it. It just always seems to be that error > whenever > >> auto-landmarks fails. > >> > >> In your case, I can't be sure what caused it to fail. It could > be that > >> your surface isn't in 711-2B stereotaxic space. Eventually, > we'd like > >> to add handling to make the process of getting surfaces into > 711-2B more > >> seemless. But we're not there yet. > >> > >> But really you're going to have to improve the quality of the > >> segmentation that generated this surface before registeration makes > >> sense. > >> > > >> > How does the automatic paint generation of the core6 > landmark, and > >> > other sulci labels actually work? By my understanding, Caret, > once it > >> > finds the sylvian fissure correctly, should be able to locate the > >> > central sulcus shouldn't it? and after that it should find > all the > >> > other sulci paints by using the B12 atlas. > >> This isn't really documented/published anywhere, as far as I > know. I'll > >> give you my oversimplified take on it: The surface must be in > 711-2B, > >> so that the probabilistic volumes in your > >> $CARET_HOME/data_files/sulcal_identification directory align > properly > >> with it. Several ROI volumes (e.g., CeS, HF, SF) are mapped to the > >> surface, thereby forming surface ROIs, which are intersected with > >> thresholded Depth metrics to pinpoint the location of key > anatomical > >> structures. Borders are drawn and trimmed to our Core 6 landmark > >> specifications. > >> > thank you all for any suggestion/input/critique! > >> > > >> > > >> > Cheers, > >> > > >> > Danny Kim > >> > Brainmapping > >> > Dept. of Neurology > >> > BC Children's Hospital > >> > Vancouver, BC > _______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
