Thank you all for the very useful feedback! A quick question before I address the other points: I used the average surface off all controls in one image, and average surface of all patients in the other. If I understood Anderson's comment correctly, it'd be better to use exactly the same inflated surface for both? What should be the underlay then? The inflated surface of the entire sample (patients + controls), even though the myelin maps are only obtained on parts of that total sample? Best, Darko
On Thu, Aug 9, 2018 at 12:14 AM, Winkler, Anderson (NIH/NIMH) [E] < [email protected]> wrote: > Hi Darko, hi all, > > > > Are these average images using the exact same color scale? It’d help if > the surface geometries were the same (e.g., the same inflated for all). But > assuming that the color scale is the same, and trying to see over the > geometry differences, looks like the averages between the two groups aren’t > that massively different, or at least not that asymmetric, correct? > > > > Then taking this together with the Cohen’s d maps, for me this raises the > possibility that there could be a large outlier in one hemisphere – perhaps > a subject who moved a bit too much laterally between collection of T1 and > T2? Or maybe a subject whose right pial reconstruction wasn’t correct and > led to a much thinner cortex, thus with the myelin profile taken too far > away from the others (i.e., the mid-thickness too off)? > > > > I guess you might want to investigate on a per-subject and per vertex > basis... For example, take some vertices where the difference between L and > R is large and check how the values vary between subjects, see if there are > too influential observations, then open the images for this/these subjects > and see if something went wrong along the pipeline… You can also open the > residuals and see if there are large outliers with a lot of spatial spread. > > > > These are just suggestions… hope they could help anyway… > > > > All the best, > > > > Anderson > > > > > > *From: *"Glasser, Matthew" <[email protected]> > *Date: *Wednesday, August 8, 2018 at 17:05 > *To: *"Harms, Michael" <[email protected]>, Darko Komnenić < > [email protected]> > > *Cc: *"Winkler, Anderson (NIH/NIMH) [E]" <[email protected]>, " > [email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > Also please post the T1w and T2w image protocols and make sure to use the > second scaling option for the percentiles (which takes percentiles of the > absolute value so that positive and negative colors mean the same > thing—this probably should be the default. > > > > Matt. > > > > *From: *"Harms, Michael" <[email protected]> > *Date: *Wednesday, August 8, 2018 at 2:13 PM > *To: *Darko Komnenić <[email protected]>, Matt Glasser < > [email protected]> > *Cc: *"Winkler, Anderson (NIH/NIMH) [E]" <[email protected]>, " > [email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > > > Hi, > > Viewing on the inflated surface would probably be more helpful. > > > > Also, how many subjects per group? > > > > Cheers, > > -MH > > > > -- > > Michael Harms, Ph.D. > > ----------------------------------------------------------- > > Associate Professor of Psychiatry > > Washington University School of Medicine > > Department of Psychiatry, Box 8134 > > 660 South Euclid Ave > <https://maps.google.com/?q=660+South+Euclid+Ave&entry=gmail&source=g>. > Tel: 314-747-6173 > > St. Louis, MO 63110 Email: [email protected] > > *From: *Darko Komnenić <[email protected]> > *Date: *Wednesday, August 8, 2018 at 2:08 PM > *To: *"Glasser, Matthew" <[email protected]> > *Cc: *"Harms, Michael" <[email protected]>, "Winkler, Anderson (NIH/NIMH) > [E]" <[email protected]>, "[email protected]" < > [email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > Thanks a lot for the clarifications everyone! > > The first two images attached are average myelin maps for controls and > patients, respectively. > > The third image is what I hope to be the effect size map. I added the > -saveglm flag to palm, and then used command -cifti-create-dense-from-template > to merge the dpv_cohen files for left and right hemisphere. Let me know if > this does not work. > > Thanks in advance for any comments! > > Best, > > Darko > > > > > > > > On Wed, Aug 8, 2018 at 2:22 AM, Glasser, Matthew <[email protected]> > wrote: > > What is an effect size map? An effect size map is in the case of a t-test > of difference between two groups is simply the difference in the means. A > standardized effect size maps is a Cohen’s d, which is helpful if you want > to compare effect sizes of different measures that are not on the same > scale. Looking at both the difference between means and the means > themselves could be helpful in tracking down artifacts. Looking at maps of > statistical significance is not helpful, despite what you see typically > done in the neuroimaging literature. > > > > Matt. > > > > *From: *<[email protected]> on behalf of "Harms, > Michael" <[email protected]> > *Date: *Tuesday, August 7, 2018 at 3:47 PM > *To: *"Winkler, Anderson (NIH/NIMH) [E]" <[email protected]>, > Darko Komnenić <[email protected]> > > > *Cc: *"[email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > > > A Cohen’s d map of the group difference might be helpful, but even more > basic would be to just merge the individual myelin maps (-cifti-merge, or > use wb_shortcuts -cifti-concatenate) and then average them (-cifti-average). > > > > Cheers, > > -MH > > > > -- > > Michael Harms, Ph.D. > > ----------------------------------------------------------- > > Associate Professor of Psychiatry > > Washington University School of Medicine > > Department of Psychiatry, Box 8134 > > 660 South Euclid Ave > <https://maps.google.com/?q=660+South+Euclid+Ave&entry=gmail&source=g>. > Tel: 314-747-6173 > > St. Louis, MO 63110 Email: [email protected] > > *From: *"Winkler, Anderson (NIH/NIMH) [E]" <[email protected]> > *Date: *Tuesday, August 7, 2018 at 3:02 PM > *To: *Darko Komnenić <[email protected]>, "Harms, Michael" < > [email protected]> > *Cc: *"[email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > Hi Darko, > > > > The option “-saveglm” in PALM should produce Cohen’s d maps. > > > > All the best, > > > > Anderson > > > > > > *From: *Darko Komnenić <[email protected]> > *Date: *Tuesday, August 7, 2018 at 15:25 > *To: *"Harms, Michael" <[email protected]> > *Cc: *"[email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > Hi Michael, > > I don't seem to have an average myelin map as an output of the analysis. > Is it just something I can make by merging and averaging individual myelin > maps in workbench, or should it have been the output of the GLM analysis? > Sorry for this most likely really basic question. > > Thanks in advance! > > Best, > > Darko > > > > On Tue, Aug 7, 2018 at 3:53 PM, Harms, Michael <[email protected]> wrote: > > > > It might be helpful to simply see the average myelin map for each group. > Do those look appropriate? > > > > Cheers, > > -MH > > > > -- > > Michael Harms, Ph.D. > > ----------------------------------------------------------- > > Associate Professor of Psychiatry > > Washington University School of Medicine > > Department of Psychiatry, Box 8134 > > 660 South Euclid Ave > <https://maps.google.com/?q=660+South+Euclid+Ave&entry=gmail&source=g>. > Tel: 314-747-6173 > > St. Louis, MO 63110 Email: [email protected] > > *From: *<[email protected]> on behalf of Darko > Komnenić <[email protected]> > *Date: *Tuesday, August 7, 2018 at 6:50 AM > *To: *"Glasser, Matthew" <[email protected]> > > > *Cc: *"[email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > Hi Matt, > > I'm not sure what you mean by a simple effect size map. The file overlaid > on the surfaces here is a file showing the difference between control and > patient groups, after correction for multiple comparisons. Do I use that to > make an effect size map, or something else? > > Best, > > Darko > > > > On Mon, Aug 6, 2018 at 9:03 PM, Glasser, Matthew <[email protected]> > wrote: > > Can you just make a simple effect size map with + and – and no > thresholding? > > > > Matt. > > > > *From: *Darko Komnenić <[email protected]> > *Date: *Monday, August 6, 2018 at 12:09 PM > > > *To: *Matt Glasser <[email protected]> > *Cc: *"[email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > Hi Matt, > > here's a screenshot of an unthresholded map. > > Thanks in advance! > > Best, > > Darko > > > > On Fri, Aug 3, 2018 at 10:12 PM, Glasser, Matthew <[email protected]> > wrote: > > Thresholded maps of statistical significance are essentially > uninterpretable as to the existence of artifacts. Please provide an > unthresholded effect size map. > > > > Peace, > > > > Matt. > > > > *From: *Darko Komnenić <[email protected]> > *Date: *Friday, August 3, 2018 at 12:12 PM > *To: *Matt Glasser <[email protected]> > *Cc: *"[email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Finding a significant difference only in one > hemisphere > > > > Hi Matt, > > here's a screenshot > > Thanks in advance! > > Best, > > Darko > > > > On Thu, Aug 2, 2018 at 10:01 PM, Glasser, Matthew <[email protected]> > wrote: > > How about posting some pics. > > > > Matt. > > > > *From: *<[email protected]> on behalf of Darko > Komnenić <[email protected]> > *Date: *Thursday, August 2, 2018 at 9:16 AM > *To: *"[email protected]" <[email protected]> > *Subject: *[HCP-Users] Finding a significant difference only in one > hemisphere > > > > Dear HCP experts, > > I ran an analysis comparing cortical myelination between patients and > controls and found a significant difference between groups, but only in the > left hemisphere. What's making me doubt my results is that the differences > are observed in large areas of the left hemisphere, but absolutely nowhere > in the right, and we have no theoretical reasons to expect such a > lateralization. > > Do you have any ideas what might have gone wrong? I have 17 patients and > 25 controls, is this maybe too small of a sample? Should I double check > certain analysis steps? Basically any ideas are welcome, since I really > cannot interpret these findings at all, it seems. > > Thanks in advance! > > Best, > > Darko > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
