[RESOLVED] Issue with 'hcp_diffusion'

Use Questions
1

Hi,
I’m trying to use qunex for preprocessing with T1, T2, rs-fMRI, and DWI data. This time I ruled out data and GPU issues, but I encountered another error.An error also occurred during the run_eddy step in hcp_diffusion . The log file seems to indicate “Segmentation violation, Address not mapped, Offending address = 0xb8”, but i don’t know how to solve it. Below is my log file.
runlog:

Log-hcp_diffusion-2025-08-16_20.25.42.447789.log (3.0 KB)

comlog:

error_hcp_diffusion_STG005_2025-08-16_20.25.42.466932.log (119.1 KB)

Thanks a lot for all!

Best,

Kelly

2

Hi Kelly,

Do you get the same error with other sessions as well, or is this popping up with this session only?

I see you are adding --data_is_shelled, is the data indeed shelled?

As a sanity check, you could also try without a GPU as the error happens during eddy GPU processing.

Best, Jure

3

Hi, Jure

I’m so sorry for the late reply.

Other sessions didn’t have this error. According your suggestion, I have checked the bvalue and bvector file to ensure that the data is shelled. I found that all the numbers in the file are zero. So we started troubleshooting errors from the data transmission stage again, and the error is highly likely to occur during the data export stage. I will let you know if there is any progress.

Thanks for your kindly help！

Best, Kelly

4

Hi, thanks for the update. If you need any additional assitance down the line, feel free to post here.

Best, Jure

5

Hi, Jure

I still have a slightly confused question.

seqtype='t1_mprage_sag_iso t2_space_sag ep2d_se_AP ep2d_se_PA Rest_fMRI_ep2d_bold_AP_488 Rest_fMRI_ep2d_bold_PA_488 task_fMRI_ep2d_bold_AP_455 task_fMRI_ep2d_bold_PA_455 ep2d_se_AP ep2d_se_PA ep2d_diff_mddw_99_AP ep2d_diff_mddw_99_PA'

seqtype=${seqtype// /\#} # neeed to use # as delimiter

seqname='T1w T2w SE-FM-AP SE-FM-PA bold:rest:phenc(AP) bold:rest:phenc(PA) bold:task:phenc(AP) bold:task:phenc(PA) SE-FM-AP SE-FM-PA DWI:dir99_AP DWI:dir99_PA'

seqname=${seqname// /\#} # neeed to use # as delimiter

There are two types of fmri data for the same subject in my dataset: rest-fmri and task-fmri. I would like to preprocess and extract features using Qunex for both rest-fmri and task-fmri together, so that subsequent data processing can be standardized. rest-fmri contains BOLD1 in the AP direction and BOLD2 in the PA direction, while task fmri contains BOLD3 in the AP direction and BOLD4 in the PA direction. I am now adding “-- hcp_icafix_bars=” rest: BOLD_1, BOLD_2 | task: BOLD_3, BOLD_4 "" in hcp_icafix and hcp_post_fix to separate the rest and task folders in “ssd2/qzheng/STG.all/qunex/STG.all/sessions/STG005/hcp/STG005/MNiNonLinear/Results”. In hcp_msmall, only rest-fmri was processed using “-- hcp_icafix_bones=” rest: BOLD_1, BOLD_2 "".
And is it correct for me to use this command when going to map_hcp_data?

   map_hcp)
    qunex_container map_hcp_data \
        --sessionsfolder="${WORK_DIR}/${STUDY_NAME}/sessions" \
		    --batchfile="${WORK_DIR}/${STUDY_NAME}/processing/batch.txt" \
        --sessions="${SUBJ}" \
        --hcp_cifti_tail=_Atlas \
        --bolds=all \
        --additional_bolds=rfMRI_REST \
        --bind="${SDIR}:${SDIR}"  \
        --container="${QUNEX_CONTAINER}";
    ;;

The result returned four folders: BOLD1, BOLD2, BOLD3, and BOLD4. So I am a bit confused now whether the simultaneous processing of two types of fmri data needs to be separated in the hcp_icafix stage, or whether BOLD1, BOLD2, BOLD3, and BOLD4 should always be used for processing until only task analysis is performed on BOLD3 and BOLD4 during the steps of create_conc and so on.

I would like to know if it is feasible to handle both rest and task simultaneously, and if my current approach is correct. If it is not right, how should I modify it ?

Best, Kelly

6

Hm, I am not sure as the datasets I worked with so far had only resting state data. I think the best way to go here would be to ask on the HCP Pipelines forum (https://groups.google.com/a/humanconnectome.org/g/hcp-users) about how to run HCP ICAFix and MSMall if you have 2 REST bolds and 2 TASK bolds. Once we get that info, I can help you set everything up with QuNex.

Best, Jure