Hello Ryota, Please see my comments in-line below.
Thanks, --Greg ____________________________________________________________________ Greg Burgess, Ph.D. Staff Scientist, Human Connectome Project Washington University School of Medicine Department of Anatomy and Neurobiology Phone: 314-362-7864 Email: [email protected] Ryota Tomioka wrote: > Hi Greg, > > I was looking a bit more into other tasks and would appreciate if you > (or others) could clarify the issues below: > > Language task > I found that some times the response event starts from a time > exceeding the number of frames in a run. For example, > $cat 100307/MNINonLinear/Results/tfMRI_LANGUAGE_LR/EVs/response_math.txt > 39.371 2.965 1.0 > 51.926 2.956 1.0 > 94.908 2.965 1.0 > 107.703 2.959 1.0 > 148.806 2.968 1.0 > 161.614 2.959 1.0 > 171.73 2.961 1.0 > 184.138 2.966 1.0 > 229.213 2.964 1.0 > > Here the number of frames is 316 (=227.52 s), whereas the last > response event happens at 229.213 s. The RL phase encoding also had > the same issue. For LANGUAGE, the event-related EVs are actually part of the larger task structure. In other words, each story or math trial consists of a presentation, a question, and a response. The "dummy" math events will start if the scan has not completed, and those events will continue for a full trial (present, question, and response events). The scan could end somewhere in the middle of the trial, in which case the response event might start after the scan has completed. > > Emotion processing task > There seem to be only fear.txt and neut.txt corresponding to the face > blocks and shape blocks. How can I see if the face shown was an angry > or fearful face? How about the subjects' responses? If you have contrasts of interest that are not included in the EVs provided by HCP, you can parse the TAB.txt files in each EV folder to create your own EVs. The documentation at http://humanconnectome.org/documentation/S500/HCP_S500_Release_Appendix_VI.pdf will provide some additional information to help you parse those files. > Also similarity to > the language task, the last block exceeds the number of frames. Is > this is the bug explained in the reference manual? Does this affect > all 500 subjects? Yes, this is due to the bug described in the reference manual. To maintain consistency across participants, we have not changed the script mid-stream. Therefore, it does affect all 500 subjects. > > > Thanks, > Ryota > > > > On Thu, Sep 4, 2014 at 6:42 PM, Greg Burgess<[email protected]> wrote: >> Hi Ryota, >> >> Please see my comments inline below. I hope the additional info helps! >> >> --Greg >> >> ____________________________________________________________________ >> Greg Burgess, Ph.D. >> Staff Scientist, Human Connectome Project >> Washington University School of Medicine >> Department of Anatomy and Neurobiology >> Phone: 314-362-7864 >> Email: [email protected] >> >> On Sep 2, 2014, at 7:32 AM, Ryota Tomioka<[email protected]> wrote: >> >>> Hi all, >>> >>> I have been looking into the task fMRI data and comparing the timings >>> saved in EVs folder with those explained Table 4 in Barch et al. >>> (2013) "Function in the human connectome: Task-fMRI and individual >>> differences in behavior". >>> >>> I am new to fMRI, so maybe I am missing something basic but the >>> numbers do not seem to exactly match what I expect. I would greatly >>> appreciate if some of you could l clarify this for me. >>> >>> 1. What is the length of a frame? >>> If I use read_avw.m function provided by FSL, I get 0.72 s/frame. >> The TR is definitely 0.720 s per frame. >> >>> The >>> numbers I get from Table 4 in Barch et al. seem to vary from a task to >>> another. For example, for WM task, 5:01 / 405 frames = 0.74 s/frame. >> Our apologies for this confusion. The “run duration” values were read >> directly from the Siemens syngo console. That "run duration" includes >> additional preparation volumes that are present before the beginning of the >> actual task scans. More specifically, the multiband fMRI sequence used in >> HCP collects eight MB8 scans that are re-combined to make the SBRef image >> that is necessary for registration and MB reconstruction, as well as five >> “dummy” scans that are discarded to allow the MR signal to reach steady >> state. Those thirteen volumes add 9.36 s of preparation time before the task >> scans. >> >> If you take the number of frames per run multiplied by 0.720 s per frame, >> and add 9.36 s for the preparation volumes, your sum will be within 500 ms >> (rounding error) of the run duration reported in Table 4. >> >> >>> Actually, the numbers in this table does not add up as I would expect. >>> For example, for the motor task, according to the table the run >>> duration is 3:34 (min). But 10 * (12 + 3) + 3 * 15 = 3:15 (min). >> I would suggest ignoring the preparation volumes and instead accounting for >> the number of frames * the TR: (284*.720) = 204.48 s for the MOTOR task. In >> this example, you’ve forgotten to add the 8 s “task initiation countdown” >> that occurs at the start of the run for most tasks (see last row of Table >> 4). Therefore, the actual duration of the task is the sum of the task >> blocks, the fixation blocks, and the initial countdown: [10 * (12 + 3)] + >> [3 * 15] + [8] = 203 s. >> >> Because there is some variability in the timing of events in E-Prime due to >> the use of the Windows OS, we have padded the end of all task runs with a >> small number of additional volumes to ensure that we are acquiring BOLD >> signal throughout the entire task, even if the task events were delayed due >> to Windows / E-Prime. This is also the reason why precise timing is provided >> for each participant, rather than using a general set of EVs across all >> participants. >> >> >>> 2. Does the first fMRI frame corresponds to the origin (0s) of the >>> times in *.txt files in EVs folder? >>> The first trial (or cue) saved in EVs folder seems to be always around >>> 8s and the last trial seems to end always much earlier than the >>> duration of a run. I was wondering what time the timings in *.txt >>> files are measured from. >> The first fMRI frame (after the preparation volumes) corresponds to the >> origin (0 s) in the .txt EV files. The first task trial occurs after an 8 s >> countdown to prepare participants. Therefore, the first event in the EV >> files occurs around 8s after the onset of the first fMRI frame. >> >>> 3. What is Sync.txt in EVs folder? >>> It doesn't seem to be explained in the reference manual. >> The variables required to convert the E-Prime timing information in the >> TAB.txt file to the timing provided in the EV files is explained in an >> Appendix to the Reference Manual: >> http://humanconnectome.org/documentation/S500/HCP_S500_Release_Appendix_VI.pdf >> >> The Sync.txt file contains the offset in ms between the start of the E-Prime >> script and the onset of the first fMRI frame / countdown event. This value >> is essentially subtracted from the timing in the TAB.txt files to provide >> the timing relative to the onset of the first fMRI frame rather than the >> start of the E-Prime script. >> >> >>> Thanks, >>> >>> Ryota Tomioka >>> _______________________________________________ >>> HCP-Users mailing list >>> [email protected] >>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
