Hi Yann, In a GLM, the baseline / intercept / B0 reflects the mean of the residual BOLD time series after removing variance explained by all other regressors. Beta weights for other conditions can be conceptualized as BOLD differences between those conditions and the baseline.
When the task includes unmodelled fixation blocks (i.e., there are no regressors in the GLM for fixation time points), the B0 intercept is the mean of the residual BOLD signal, which includes variance related to activation during the fixation time points, as well as residual variance from the other time points (the mean of this random error is expected to be 0). The beta weight for Condition X reflects the difference between X and B0 (in other words, X versus baseline). When the task does not include unmodelled fixation blocks (i.e., the task regressors include all time points in the scan), the B0 intercept is STILL the mean of the residual BOLD signal. But, in this situation, B0 only reflects the mean of the residual variance after accounting for the task. B0 reflects the mean of the random error, which is expected to be 0. The first problem in interpreting Condition X versus baseline when there are no unmodelled time points is that you have no idea what B0 is in terms of brain activation. B0 is specifically the mean signal from the scanner after accounting for everything that you know about brain activity. The second problem is that you expect B0 to equal 0, but the nature of random error is that the mean of the residual variance could be anything. If you’ve failed to model the HRF of X correctly in that region, or if there is any scanner artifact, the mean of that variance could be non-zero. What you need to do for these tasks is to focus on contrasts between conditions. The contrast X-Y is equal to (X-B0)-(Y-B0). The B0 terms cancel out, and you know what X and Y mean with respect to brain activation. So, those contrasts between conditions can be interpreted in terms of brain activation, even though the contrast of each condition versus “baseline” cannot be interpreted in terms of brain activation. --Greg ____________________________________________________________________ Greg Burgess, Ph.D. Staff Scientist, Human Connectome Project Washington University School of Medicine Department of Psychiatry Phone: 314-362-7864 Email: [email protected]<mailto:[email protected]> On May 4, 2018, at 8:43 AM, Yann Le Guen <[email protected]<mailto:[email protected]>> wrote: Dear HCP community, I dug back in an old thread because it seems the point 2. was not answered. For the Language tasks, there are no fixation blocks so it seems that the baseline is indeed implicit. I remember reading somewhere that it was the intercept GLM ? However, could you clarify this and what it means in term of brain activations ? Best regards, Yann 2016-10-18 15:50 GMT+02:00 Harms, Michael <[email protected]<mailto:[email protected]>>: Hi, The fixation blocks establish the implicit baseline. There is no beta map for fixation. cheers, -MH -- Michael Harms, Ph.D. ----------------------------------------------------------- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave. Tel: 314-747-6173 St. Louis, MO 63110 Email: [email protected]<mailto:[email protected]> From: Vadim Axelrod <[email protected]<mailto:[email protected]>> Date: Tuesday, October 18, 2016 at 8:12 AM To: Michael Harms <[email protected]<mailto:[email protected]>> Cc: "Glasser, Matthew" <[email protected]<mailto:[email protected]>>, "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Re: [HCP-Users] Extracting beta values from analyzed task-data Hi Micahel and all, I am sorry to bother you again, but as a follow-up of you last remark, I feel that I am confused about the way the baseline is treated: 1. How contrast and beta map can be the same thing? For example, let's take Social cognition exp. At the first (or second) level there should be 3 regressors: RANDOM, TOM and fixation. So, each one of them should have separate beta map. So, the contrast between RANDOM and fixation is the difference between the beta maps of RANDOM and fixation. But what I see is that the resultant difference map is equivalent to beta map of RANDOM. To best of my understanding, fixation is a condition as any other condition. 2. Is the baseline reflects fixation blocks? If so, in Language exp. there is no Fixation block and in Emotional exp. there is only one fixation block at the end. So, what does the contrast Story > baseline reflect? 3. I wanted to find beta estimate for the fixation in Social cognition exp. The only possible pe file for the fixation is the pe file in cope3.feat (RANDOM-TOM), because two others are the betas of the RANDOM and TOM. However, its map almost perfectly correlates with the contrast map of RANDOM-TOM. So, is there beta map for fixation blocks? Thanks a lot, Vadim On Mon, Oct 17, 2016 at 10:30 PM, Harms, Michael <[email protected]<mailto:[email protected]>> wrote: And, to answer the Q in your 1st paragraph: yes. BTW: Since you want the “betas", you should use the pe1.dtseries.nii files. But, as an aside, note that these pe’s/cope’s are the results of the Level2 analysis, essentially averaging the Level 1 (individual runs). Because all the contrasts were defined in Level 1, and then essentially averaged for Level 2, I’m pretty sure that within each GrayordinateStats/cope{i}.feat directory, the actual grayordinate-values of the pe1.dtseries.nii and cope1.dtseries.nii files are identical. (i.e., we didn’t define any “new” contrasts specifically at Level 2, which means that the pe1 and cope1 will be the same within each contrast). cheers, -MH -- Michael Harms, Ph.D. ----------------------------------------------------------- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave.Tel: 314-747-6173<tel:314-747-6173> St. Louis, MO 63110Email: [email protected]<mailto:[email protected]> From: "Glasser, Matthew" <[email protected]<mailto:[email protected]>> Date: Monday, October 17, 2016 at 2:05 PM To: Vadim Axelrod <[email protected]<mailto:[email protected]>>, Michael Harms <[email protected]<mailto:[email protected]>> Cc: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Re: [HCP-Users] Extracting beta values from analyzed task-data Grand mean 10000. Find out what SPM uses and adjust your internal scale accordingly... Peace, Matt. ________________________________ From: Vadim Axelrod <[email protected]<mailto:[email protected]>> Sent: Monday, October 17, 2016 1:41:35 PM To: Harms, Michael Cc: Glasser, Matthew; [email protected]<mailto:[email protected]> Subject: Re: [HCP-Users] Extracting beta values from analyzed task-data Thank you for the help! So, if I want to get a beta estimate for the condition, I should take pe image in the cope#.feat folder, where this condition is compared vs. fixation. For example, in the Language task, for the Math beta I should take pe image in cope1.feat (MATH-fixation) and the Story betas are stored in cope2.feat (STORY-fixation contrast). Correct? One more question: I noticed that betas might have high values (e.g., 50). I am used to work with SPM, and their the beta estimates are order of magnitude smaller. Is there some scaling here? Harms, Michael <[email protected]<mailto:[email protected]>> wrote: No, in the updated pipeline, we simply have some code to merge the copes/betas from the different contrasts into a single file for each task (as we did for the zstat maps, which perhaps somewhat confusingly, are the files named as <subject>_tfMRI_<task>_level2_hp200_s?.dscalar.nii or <subject>_tfMRI_<task>_level2_hp200_s?_MSMAll.dscalar.nii — i.e., “zstat” is not part of the file name). There is no need to re-run the pipelines. You can just use the appropriate “merge” command in wb_shortcuts to merge the copes/betas into a similar single file. As Matt pointed out, those copes/betas are in the GrayordinateStats/cope{i}.feat/{pe1,cope1}.dtseries.nii files — “pe1” for the betas; “cope1” for the contrasts). cheers, -MH -- Michael Harms, Ph.D. ----------------------------------------------------------- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave.Tel: 314-747-6173<tel:314-747-6173> St. Louis, MO 63110Email: [email protected]<mailto:[email protected]> From: <[email protected]<mailto:[email protected]>> on behalf of Vadim Axelrod <[email protected]<mailto:[email protected]>> Date: Monday, October 17, 2016 at 8:28 AM To: "Glasser, Matthew" <[email protected]<mailto:[email protected]>> Cc: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Re: [HCP-Users] Extracting beta values from analyzed task-data Thank you, Matt. So, you mean that in the updated pipeline cope1.dtseries.nii contains multiple beta maps (two or more, depending on a contrast)? Practically, do you think in the foreseeable future there will be an update of task data for the download, that will include the betas? If not, then I should probably run the pipeline myself. Can you please advice the efficient way of doing this (see my original post). Many thanks! On Mon, Oct 17, 2016 at 4:02 PM, Glasser, Matthew <[email protected]<mailto:[email protected]>> wrote: It looks like the released HCP task fMRI data were produced using an older version of the Task analysis pipeline that does not automatically collate the beta maps into a labeled and easy to use dscalar file. The beta maps will be in these files: ${StudyFolder}/${Subject}/MNINonLinear/Results/${fMRIName}/${fMRIName}_hp200_s2_level2_MSMAll.feat/GrayordinatesStats/cope${i}.feat/cope1.dtseries.nii where ${i} = 1 to N contrasts Peace, Matt. From: <[email protected]<mailto:[email protected]>> on behalf of Vadim Axelrod <[email protected]<mailto:[email protected]>> Date: Sunday, October 16, 2016 at 7:31 AM To: Stephen Smith <[email protected]<mailto:[email protected]>> Cc: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Re: [HCP-Users] Extracting beta values from analyzed task-data Thank you Stephen for the prompt reply. But why there is only one pe1 file with only one volume in it? For the contrast like, story > math I would expect to find two beta images, one for the "story" and one for the "math". In addition, the pe1.dtseries.nii maps are very correlated with tstat1.dtseries.nii (R>0.85) and there is similar correlation for all contrasts of the same task/subject. On Sun, Oct 16, 2016 at 2:09 PM, Stephen Smith <[email protected]<mailto:[email protected]>> wrote: HI - "pe" are the beta files and "cope" are contrasts of these. Cheers. On 16 Oct 2016, at 12:03, Vadim Axelrod <[email protected]<mailto:[email protected]>> wrote: Dear list, I downloaded the analyzed 3T HCP fMRI task-data. In the cope* subfolders I can find the outputs of the contrast calculation, e.g., Zstat1.dtseries.nii. However, I do not see beta estimates for each regressor. Do I miss something or they are just not included in the analyzed archive. If the latter is correct, should I download non-analyzed data and run the analysis myself? What is the simplest way to to that? The only instructions that I found are from the last year course, and I am not sure how valid they are: https://www.humanconnectome.org/courses/2015/exploring-the-human-connectome.php https://wustl.box.com/shared/static/0vcsqc0yg3ncqmirryyxpgsiqhxww659.pdf Probably someone can post more recent instructions from the last month course. Many thanks for the help, Vadim _______________________________________________ HCP-Users mailing list [email protected]<mailto:[email protected]> http://lists.humanconnectome.org/mailman/listinfo/hcp-users --------------------------------------------------------------------------- Stephen M. Smith, Professor of Biomedical Engineering Head of Analysis, Oxford University FMRIB Centre FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK +44 (0) 1865 222726<tel:%2B44%20%280%29%201865%20222726> (fax 222717) [email protected]<mailto:[email protected]> http://www.fmrib.ox.ac.uk/~steve --------------------------------------------------------------------------- Stop the cultural destruction of Tibet<http://smithinks.net/> _______________________________________________ HCP-Users mailing list [email protected]<mailto:[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]<mailto:[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]<mailto:[email protected]> http://lists.humanconnectome.org/mailman/listinfo/hcp-users _______________________________________________ HCP-Users mailing list [email protected]<mailto:[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
