The WU-Minn Human Connectome Project (HCP) releases imaging data collected on subjects on a regular basis. Details on what data is included in each release are available in the Reference Manual for that release on the HCP documentation page.  On this wiki page, we document known updates, data issues, bugs, and known missing files in the data distributed with each release (including those distributed in Connectome in a box for each release) and plans to fix or patch each bug. Date added is listed at the beginning of entries.

1200 Subjects Release (March 1, 2017)

Major Issues 

3T Functional MRI data from a batch of 25 subjects were incorrectly processed, with all "RL" runs (right-left phase encoding) being processed with the wrong phase encoding direction. Not only are the preprocessing results for the RL functional runs themselves incorrect, but the MSM-All registration for these subjects, which is dependent on the correct preprocessing outputs, is also incorrect.  Therefore any analyses using the MSM-All aligned data from these subjects would be affected. Further, 7 of these subjects also have 7T data, so analyses of 7T data that use MSM-All aligned data would be affected as well (since the MSM-All registration applied to the 7T data was based on the 3T data). No Test-Retest subjects were affected. 

The 3T "LR" (left to right phase encoding) runs for these subjects were processed correctly, but we recommend using both runs in analyses to mitigate phase encoding bias. The MSM-Sulc data for the 3T LR functional runs and 7T functional data are not misaligned, but we don't recommend using MSM-Sulc aligned data for cross subject analyses.

Recommendations:

Due to the incorrect preprocessing of half the runs, we recommend excluding these subjects from any further analyses of the 3T functional data

7T preprocessed Functional data (all runs) were processed correctly and can be used. However, due to the non-optimally aligned MSM-All surfaces, if you are performing cross-subject analyses using the 7T functional data, your options are to either: (1) exclude these subjects from your analysis, (2) confirm that the MSM-All registration in these subjects is still reasonable*, or (3) use the MSM-Sulc registered surfaces instead for all subjects (not recommended).

3T and 7T preprocessed Diffusion data for these subjects was not affected by the functional preprocessing error and can be used. However, due to the non-optimally aligned MSM-All surfaces, if you are performing cross-subject diffusion analyses involving surfaces, your options are to either: (1) exclude these subjects from your analysis, (2) confirm that the MSM-All registration in these subjects is still reasonable*, or (3) use the MSM-Sulc registered surfaces instead for all subjects (not recommended).

*One may determine that the MSM-All registration is reasonable in the affected subjects by:

- Comparing affected subjects' MSMAll myelin maps ${StudyFolder}/${Subject}/MNINonLinear/${Subject}.MyelinMap_BC_MSMAll.164k_fs_LR.dscalar to the group template available for download in https://github.com/Washington-University/HCPpipelines/blob/master/global/templates/MSMAll/Q1-Q6_RelatedParcellation210.MyelinMap_BC_MSMAll_2_d41_WRN_DeDrift.32k_fs_LR.dscalar.nii

- Looking at the affected subjects' MSMAll areal distortion maps ${StudyFolder}/${Subject}/MNINonLinear/${Subject}.ArealDistortion_MSMAll.164k_fs_LR.dscalar.nii

Subject 168139 has an anatomical anomaly, as noted in HCP Subjects with Identified Quality Control Issues (QC_Issue measure codes explained), a large posterior fossa subarachnoid cyst. The cyst was large enough in this case to cause the EPI to T1w alignment to fail severely in the fMRI Volume part of functional preprocessing, and causes all subsequent preprocessing results to be unusable. The following runs in the released data are affected and should not be used

tfMRI_EMOTION_LR

tfMRI_GAMBLING_LR

tfMRI_GAMBLING_RL

tfMRI_LANGUAGE_LR

tfMRI_LANGUAGE_RL

tfMRI_MOTOR_LR

tfMRI_MOTOR_RL

tfMRI_RELATIONAL_LR

tfMRI_RELATIONAL_RL

tfMRI_SOCIAL_LR

tfMRI_SOCIAL_RL

tfMRI_WM_LR

tfMRI_WM_RL

Data from the remaining fMRI runs for this subject (all resting state and tfMRI_EMOTION_RL) are usable. 

No planned Fix: These data are released and we do not plan to remove them from ConnectomeDB or AWS S3. We recommend excluding these runs from your analyses.

For a variety of reasons, some fMRI runs for particular subjects were released in the S1200 data release, but should not be used. The following runs are affected and should not be used

196952     tfMRI_WM_LR

748662     tfMRI_SOCIAL_RL

809252     tfMRI_SOCIAL_RL 

Data from the remaining fMRI runs for these subjects (all resting state and other tfMRI runs) are usable. 

No planned Fix: These data are released and we do not plan to remove them from ConnectomeDB or AWS S3. We recommend excluding these runs from your analyses.

The following released runs display incomplete brain coverage due to misplaced acquisition Field of View (FOV). We recommend exclusion of these data from analyses.

144428          tfMRI_MOTOR_RL

144428          tfMRI_GAMBLING_LR

144428          tfMRI_MOTOR_LR

168139          tfMRI_EMOTION_LR

186545          tfMRI_SOCIAL_LR

186545          tfMRI_SOCIAL_RL

186545          tfMRI_RELATIONAL_LR

186545          tfMRI_RELATIONAL_RL

186545          tfMRI_EMOTION_LR

186545          tfMRI_EMOTION_RL

192237          tfMRI_MOTOR_RL

223929          tfMRI_RELATIONAL_LR

320826          tfMRI_EMOTION_RL

644044          tfMRI_RELATIONAL_LR

644044          tfMRI_RELATIONAL_RL

644044          tfMRI_EMOTION_LR

644044          tfMRI_EMOTION_RL

822244          tfMRI_GAMBLING_LR

822244          tfMRI_MOTOR_RL

870861          tfMRI_GAMBLING_LR

870861          tfMRI_MOTOR_LR

870861          tfMRI_MOTOR_RL

947668          tfMRI_GAMBLING_LR

947668          tfMRI_GAMBLING_RL

947668          tfMRI_MOTOR_RL

Data from the remaining fMRI runs for these subjects (all resting state and other tfMRI runs) are usable. 

No planned Fix: These data are released and we do not plan to remove them from ConnectomeDB or AWS S3. We recommend excluding these runs from your analyses.

Age and Gender-adjusted T-scores for the Achenbach Self Report behavioral survey have been mislabeled/misdefined as Age and Gender adjusted Percentile scores as variable names and in Data Dictionary descriptions since these scores were released for the 500 Subjects data release. The affected scores are:

Current                           Correct

ASR_Anxd_Pct              ASR_Anxd_T 

ASR_Witd_Pct               ASR_Witd_T 

ASR_Soma_Pct             ASR_Soma_T

ASR_Thot_Pct               ASR_Thot_T

ASR_Attn_Pct                ASR_Attn_T

ASR_Aggr_Pct               ASR_Aggr_T

ASR_Rule_Pct               ASR_Rule_T

ASR_Intr_Pct                 ASR_Intr_T

DSM_Depr_Pct              DSM_Depr_T

DSM_Anxi_Pct               DSM_Anxi_T

DSM_Somp_Pct            DSM_Somp_T

DSM_Avoid_Pct             DSM_Avoid_T

DSM_Adh_Pct               DSM_Adh_T

DSM_Antis_Pct              DSM_Antis_T

Completed Fix: Labels and definitions of T-Scores for the ASR Syndrome and DSM-Oriented Scales were updated with the rerelease of the 7T HCP functional data in April 2018. 

Between HCP data collection for subjects released in the 500 Subjects Release and 900 Subjects Release, NIH Toolbox changed how the Words In Noise (WIN) audition test was administered to subjects (through computer speakers in v1 to through earphones in v2). Due to how our internal database was recording the scores for WIN v1 and WIN v2 separately, only the v1 scores have been correctly released in ConnectomeDB under the "Noise_Comp" variable, the WIN Computed score. Subjects tested with WIN v2 (most subjects newly released in S900 and S1200 releases) erroneously have either no score listed or a score of -99. NIH Toolbox has verified that the v1 and v2 WIN scoring has been normed to be directly comparable, so we plan to combine the v1/v2 scores for release as the single "Noise_Comp" variable.

Completed Fix:Words In Noise v2 scores were added to the "Noise_Comp" variable released with the rerelease of the 7T HCP functional data in April 2018. 

For many previously-released participants (specifically, subjects initially released up to and including the "500 Subjects Release"), some "In-scanner performance" EMOTION variables were swapped in behavioral data .csv files downloaded from ConnectomeDB (either from the quick downloads section of the S1200 project page or from the subject dashboard). Specifically, the Emotion_Task_Acc and Emotion_Task_Face_Acc variables were swapped, and the Emotion_Task_Median_RT and Emotion_Task_Face_Median_RT variables were swapped. The following data were not affected: .csv files for individual scan runs (obtained via Connectome-in-a-Box or downloaded .zip packages), and values for participants that were first released in the 900 Subjects Release and subsequent releases.

Completed Fix: The EMOTION variables in ConnectomeDB .csv files were corrected as of July 28, 2017. The EMOTION task accuracy and reaction time variables for all affected participants are correct in spreadsheets downloaded after this date.

A bug in the 7T fMRIVolume pipeline caused all Anterior->Posterior (AP phase-encoding direction) runs in the 7T fMRI data on all subjects to be unwarped using the incorrect direction, causing registration and all subsequent processing results to be incorrect. Since the unwarping is a very early processing step, all AP run 7T fMRI data is affected, including all preprocessed and FIX-cleaned resting state, retinotopy, and movie-watching data. All 7T fMRI data should not be used. Therefore, we have made this data unavailable for download in ConnectomeDB until fixed by reprocessing. Unprocessed 7T fMRI and all 7T diffusion MRI data is unaffected by this issue and is still available for download. 

Completed Fix: All 7T fMRI data was reprocessed with the bug fixed and rereleased in April 2018.

Two versions of the image reconstruction algorithm were used for the fMRI data (both task and resting-state) over the course of the Young-Adult HCP: version "r177" for subjects scanned early (Q1 through mid-Q3) and version "r227" for subjects scanned after that (mid-Q3 onward).  The reconstruction version has a notable signature on the fMRI data that can make a large difference in fMRI data analysis. Users wanting to use only data with the same reconstruction algorithm in their analyses can filter data on the ConnectomeDB dashboard by fMRI reconstruction version (in the Study Completion category, "fMRI_3T_ReconVrs"). See Ramifications of Image Reconstruction Version Differences for more details.  On ConnectomeDB, we also provide separate Resources for Group Average Functional Connectivity ("dense connectomes") and Parcellation+Timeseries+Netmats ("PTN") files based on both all subjects ("r177+r227") and just the "r227" subjects.

With the S1200 release, a subject data measure (QC_Issue) has been added to ConnectomeDB to flag subjects with notable issues found in the HCP Quality Control process.  The issues are notable, but were not considered severe enough to exclude the subject's imaging data from release. We are sharing these findings of issues to alert users and allow them to decide whether to include or exclude these subjects in their own analyses.  More details can be found here: HCP Subjects with Identified Quality Control Issues (QC_Issue measure codes explained)

Nineteen subjects from the S1200 Release (including two originally released in S500+MEG2 and eleven originally released in the S900 Release) do not have gradient-recalled echo (GRE) field maps accompanying their structural (T1 and T2) scans. 

Affected Subjects:

102614
111009       
111514       
115017       
121416       
130821       
138332       
179952       
299760       
300618      
392750       
406432       
429040       
633847       
662551       
679770       
688569       
693461       
815247

The field maps for these subjects were not collected or are unusable.  Field maps are not required for HCP structural processing (in contrast to fMRI processing, where field maps are required).  However, the lack of field maps prevents us from performing readout distortion correction on the structurals and will have these two effects:

Issue Cannot be Fixed

Minor Issues & Updates

The preprocessing results for the tfMRI_WM_RL run are incomplete for subject 668361, despite our collecting complete tfMRI_WM_RL data for this subject.

These directories/files are missing from the 668361/MNINonLinear/Results/tfMRI_WM_RL directory:

/EVs
/RibbonVolumeToSurfaceMapping
Movement_*.txt files
*TAB.txt files

brainmask_fs.2.nii.gz
tfMRI_WM_LR.nii.gz

tfMRI_WM_LR_Atlas.dtseries.nii
tfMRI_WM_LR_Jacobian.nii.gz
tfMRI_WM_LR_SBRef.nii.gz 
tfMRI_WM_LR_hp200_s4_level1.fsf

Planned Fix: The tfMRI_WM_RL run will be reprocessed and the missing files regenerated to update the 668361 WM tfMRI preproc package and the 668361 data in AWS S3.

Due to a late stage error in the running of the functional preprocessing pipeline for some subjects newly released with the S1200, *TAB.txt and *Physio_log.txt files are missing in the {Subject_ID}/MNINonLinear/Results/tfMRI_{TASK}_{RUN} directories for ALL tasks completed and the *Physio_log.txt files are missing in the {Subject_ID}/MNINonLinear/Results/rfMRI_{REST?}_{RUN} directory.

The missing files do exist for these subjects in {Subject_ID}/unprocessed/3T/tfMRI_{TASK}_{RUN}/LINKED_DATA/EPRIME, {Subject_ID}/unprocessed/3T/tfMRI_{TASK}_{RUN}/LINKED_DATA/PHYSIO,  and {Subject_ID}/rfMRI_{REST?}_{RUN}/LINKED_DATA/PHYSIO directories, so there was a problem with the transfer to the preprocessed directories.

Affected subjects: 

116423
120010  
121719
127226 
130114
143830 
202820
385046
401422
462139

469961
644246
723141
943862
969476
971160

Because this issue is isolated to subjects newly released with S1200, this DOES NOT affect earlier release data distributed by HCP. 

Planned Fix: The missing *TAB.txt and *Physio_log.txt files will be transferred to the appropriate preprocessing {Subject_ID}/MNINonLinear/Results directories, the affected rfMRI and tfMRI preprocessing packages will be updated in ConnectomeDB, and AWS S3 will be updated.

Initial 7T fMRI Release (June 20, 2016)

Major Issues 

A bug in the 7T fMRIVolume pipeline caused all Anterior->Posterior (AP phase-encoding direction) runs in the 7T fMRI data on all subjects to be unwarped using the incorrect direction, causing registration and all subsequent processing results to be incorrect. Since the unwarping is a very early processing step, all AP run 7T fMRI data is affected, including all preprocessed and FIX-cleaned resting state, retinotopy, and movie-watching data. All 7T fMRI data should not be used. Therefore, we have made this data unavailable for download in ConnectomeDB until fixed by reprocessing. Unprocessed 7T fMRI and all 7T diffusion MRI data is unaffected by this issue and is still available for download. 

Completed Fix: All 7T fMRI data was reprocessed with the bug fixed and rereleased in April 2018.

900 Subjects Release (December 8, 2015)

Major Issues

Age and Gender-adjusted T-scores for the Achenbach Self Report behavioral survey have been mislabeled/misdefined as Age and Gender adjusted Percentile scores as variable names and in Data Dictionary descriptions since these scores were released for the 500 Subjects data release. The affected scores are:

Current                           Correct

ASR_Anxd_Pct              ASR_Anxd_T 

ASR_Witd_Pct               ASR_Witd_T 

ASR_Soma_Pct             ASR_Soma_T

ASR_Thot_Pct               ASR_Thot_T

ASR_Attn_Pct                ASR_Attn_T

ASR_Aggr_Pct               ASR_Aggr_T

ASR_Rule_Pct               ASR_Rule_T

ASR_Intr_Pct                 ASR_Intr_T

DSM_Depr_Pct              DSM_Depr_T

DSM_Anxi_Pct               DSM_Anxi_T

DSM_Somp_Pct            DSM_Somp_T

DSM_Avoid_Pct             DSM_Avoid_T

DSM_Adh_Pct               DSM_Adh_T

DSM_Antis_Pct              DSM_Antis_T

Completed Fix: Labels and definitions of T-Scores for the ASR Syndrome and DSM-Oriented Scales were updated with the rerelease of the 7T HCP functional data in April 2018. 


Between HCP data collection for subjects released in the 500 Subjects Release and 900 Subjects Release, NIH Toolbox changed how the Words In Noise (WIN) audition test was administered to subjects (through computer speakers in v1 to through earphones in v2). Due to how our internal database was recording the scores for WIN v1 and WIN v2 separately, only the v1 scores have been correctly released in ConnectomeDB under the "Noise_Comp" variable, the WIN Computed score. Subjects tested with WIN v2 (most subjects newly released in S900 and S1200 releases) erroneously have either no score listed or a score of -99. NIH Toolbox has verified that the v1 and v2 WIN scoring has been normed to be directly comparable, so we plan to combine the v1/v2 scores for release as the single "Noise_Comp" variable.

Completed Fix: Words In Noise v2 scores were added to the "Noise_Comp" variable released with the rerelease of the 7T HCP functional data in April 2018. 

For many previously-released participants (specifically, subjects initially released up to and including the "500 Subjects Release"), some "In-scanner performance" EMOTION variables were swapped in behavioral data .csv files downloaded from ConnectomeDB (either from the quick downloads section of the S900 project page or from the subject dashboard). Specifically, the Emotion_Task_Acc and Emotion_Task_Face_Acc variables were swapped, and the Emotion_Task_Median_RT and Emotion_Task_Face_Median_RT variables were swapped. The following data were not affected: .csv files for individual scan runs (in Connectome-in-a-Box or downloaded .zip packages), and values for participants that were first released in the 900 Subjects Release and 1200 Subjects Release.

Completed Fix: The EMOTION variables in ConnectomeDB .csv files were corrected as of July 28, 2017. The EMOTION task accuracy and reaction time variables for all affected participants are correct in spreadsheets downloaded after this date.

Two subjects had some or all previously released diffusion scans which were deemed to be unusable and removed from the released data with the S1200 Release. For subject 130922, originally part of the Q3 release, the DWI_dir95_RL diffusion scan (only) was determined to be unusable. For subject150423, originally part of the Q1 release, only two diffusion scans were completed and that is not enough for useful analysis, therefore all diffusion data was removed.  

Completed Fix: As of the S1200 Release, the DWI_dir95_RL diffusion scan for subject 130922 and all diffusion data for subject 150423 were removed from the released data. The affected preprocessed dMRI data for these subjects from previous releases should not be used (whether obtained via Connectome in a Box or downloaded online).

Two versions of the image reconstruction algorithm applied to dMRI and fMRI data have been used in HCP to date: version r177 for subjects scanned in Q1 through mid-Q3, version r227 for subjects scanned mid-Q3 and after.  We were able to retroactively recon all dMRI data using the newer r227 version. However, for the fMRI data, this was not possible and data reconstructed with both r177 and r227 versions remain in ConnectomeDB. The reconstruction version makes a notable signature on the data that can make a large difference in fMRI data analysis. Users wanting to use only data with the same reconstruction algorithm applied in their analyses can filter data on the ConnectomeDB dashboard by fMRI reconstruction version (in the Study Completion category). See Ramifications of Image Reconstruction Version Differences for more details.

An artifact (characterized by a striped spatial pattern in image intensity around orbitofrontal susceptibility areas) caused by intermittent technical problems with the 32 channel head coil has been found in a limited number of fMRI scan session (24 rfMRI scans in 18 subjects; 6 tfMRI scans in 6 subjects) that were obtained during the summer of 2013. We have opted to release these data for public use with the expectation that investigators will review the data, before and after their chosen pre-processing steps, to decide whether this artifact might affect their data analysis. See  Intermittent orbitofrontal artifact affecting some fMRI scans  for details, including a list of affected subjects/scans and example images.

Minor Issues & Updates

The preprocessing results for the tfMRI_WM_RL run are incomplete for subject 668361, despite our collecting complete tfMRI_WM_RL data for this subject.
These directories/files are missing from the 668361/MNINonLinear/Results/tfMRI_WM_RL directory:

/EVs
/RibbonVolumeToSurfaceMapping
Movement_*.txt files
*TAB.txt files

brainmask_fs.2.nii.gz
tfMRI_WM_LR.nii.gz

tfMRI_WM_LR_Atlas.dtseries.nii
tfMRI_WM_LR_Jacobian.nii.gz
tfMRI_WM_LR_SBRef.nii.gz 
tfMRI_WM_LR_hp200_s4_level1.fsf

Planned Fix: The tfMRI_WM_RL run will be reprocessed and the missing files regenerated to update the 668361 WM tfMRI preproc package and the 668361 data in AWS S3.

The T1w and T2w structural data quality was reassessed and subjects 121719 & 169040 were deemed of sufficient quality to release all available imaging data with the S1200 Release.

Completed Fix: All available imaging data for subjects 121719 & 169040 was released with the S1200 Release.

Thirteen subjects from the S900 Release (including two originally released in the S500+MEG2 Release) do not have gradient-recalled echo (GRE) field maps accompanying their structural (T1 and T2) scans.

Affected Subjects:

111009       
111514       
115017       
121416       
130821              
179952              
300618      
392750       
406432       
429040       
633847       
662551       
679770              
693461

The field maps for these subjects were not collected or are unusable.  Field maps are not required for HCP structural processing (in contrast to fMRI processing, where field maps are required).  However, the lack of field maps prevents us from performing readout distortion correction on the structurals and will have these two effects:

We anticipate that a limited number of subjects that will be included in upcoming HCP releases will also have this issue. Those subjects will be identified on this Issues wiki in the appropriate release section.

500 Subjects Release (June 6, 2014) + MEG2 (November 26, 2014)

Major Issues 

Age and Gender-adjusted T-scores for the Achenbach Self Report behavioral survey have been mislabeled/misdefined as Age and Gender adjusted Percentile scores as variable names and in Data Dictionary descriptions since these scores were released for the 500 Subjects data release. The affected scores are:

Current                           Correct

ASR_Anxd_Pct              ASR_Anxd_T 

ASR_Witd_Pct               ASR_Witd_T 

ASR_Soma_Pct             ASR_Soma_T

ASR_Thot_Pct               ASR_Thot_T

ASR_Attn_Pct                ASR_Attn_T

ASR_Aggr_Pct               ASR_Aggr_T

ASR_Rule_Pct               ASR_Rule_T

ASR_Intr_Pct                 ASR_Intr_T

DSM_Depr_Pct              DSM_Depr_T

DSM_Anxi_Pct               DSM_Anxi_T

DSM_Somp_Pct            DSM_Somp_T

DSM_Avoid_Pct             DSM_Avoid_T

DSM_Adh_Pct               DSM_Adh_T

DSM_Antis_Pct              DSM_Antis_T

Completed Fix: Labels and definitions of T-Scores for the ASR Syndrome and DSM-Oriented Scales were updated with the rerelease of the 7T HCP functional data in April 2018. 

For many previously-released participants (specifically, subjects initially released up to and including the "500 Subjects Release"), some "In-scanner performance" EMOTION variables were swapped in behavioral data .csv files downloaded from ConnectomeDB (either from the quick downloads section of the S500 project page or from the subject dashboard). Specifically the Emotion_Task_Acc and Emotion_Task_Face_Acc variables were swapped, and the Emotion_Task_Median_RT and Emotion_Task_Face_Median_RT variables were swapped. The following data were not affected: .csv files for individual scan runs (in Connectome-in-a-Box or downloaded .zip packages), and values for participants  that were first released in the 900 Subjects Release and 1200 Subjects Release. (**)

(**) Technically, the small set of 15 subjects that were first released in the "S500+MEG2" Release were also not affected by this problem.

Completed Fix: The EMOTION variables in ConnectomeDB .csv files were corrected as of July 28, 2017. The EMOTION task accuracy and reaction time variables for all participants are correct in spreadsheets downloaded after this date.

An artifact (characterized by a striped spatial pattern in image intensity around orbitofrontal susceptibility areas) caused by intermittent technical problems with the 32 channel head coil has been found in a limited number of fMRI scan session (24 rfMRI scans in 18 subjects; 6 tfMRI scans in 6 subjects) that were obtained during the summer of 2013. We have opted to release these data for public use with the expectation that investigators will review the data, before and after their chosen pre-processing steps, to decide whether this artifact might affect their data analysis. See  Intermittent orbitofrontal artifact affecting some fMRI scans  for details, including a list of affected subjects/scans and example images.

Two versions of the image reconstruction algorithm applied to dMRI and fMRI data have been used in HCP to date: version r177 for subjects scanned in Q1 through mid-Q3, version r227 for subjects scanned mid-Q3 and after.  We were able to retroactively recon all dMRI data using the newer r227 version. However, for the fMRI data, this was not possible and data reconstructed with both r177 and r227 versions remain in ConnectomeDB. The reconstruction version makes a notable signature on the data that can make a large difference in fMRI data analysis. Users wanting to use only data with the same reconstruction algorithm applied in their analyses can filter data on the ConnectomeDB dashboard by fMRI reconstruction version (in the Study Completion category). See Ramifications of Image Reconstruction Version Differences for more details.

In preparation for the HCP 900 subjects data release, we identified previously unknown timing errors in the in the processed physiological log files (e.g., rfMRI_REST1_RL_Physio_log.txt). These errors have resulted in variable offsets in timing, sometimes amounting to hundreds of milliseconds. The timing errors affected rsfMRI, tfMRI, and pulse oximetry data that was collected during diffusion MRI scans. Due to limitations in computational resources, it was not possible to correct the physio files in the packages being released. Therefore, the physio log files in those packages should not be used.

Completed Fix: The corrected processed physiological log files were released for all subjects with the 900 Subjects Release in December 2015.

Subject 142626, released in the 500 Subjects Release (June 2014), has been found to have the same identity as another subject in the HCP study. Thus, we have removed all data for subject 142626 from ConnectomeDB. For any ongoing analyses, we recommend that if possible you exclude subject 142626 from your analyses.

Note: we have not removed the duplicate subject’s dataset, which is comparable in quality to other released subjects, from ConnectomeDB or from ongoing HCP analyses. However, for reasons of confidentiality, we are electing to not reveal the subject number for the duplicate subject. If the subject number of the duplicate subject is found by other investigators using HCP data, we request that you to keep this information confidential.

Subject 142626 was originally released in the 500 Subjects Release and have been available in ConnectomeDB and through Connectome in a Box (CinaB) and Amazon S3 since then.

Completed Fix: All data from subject 142626 has been removed from ConnectomeDB, CinaB, and Amazon S3 as of November 6, 2015 and will no longer be distributed by HCP.

A problem with the middle stages of the v3.1 (500 Subjects) preprocessing of diffusion data released for subject 103515 has been found. The file size for the Diffusion data for this subject is only about 400Mb, which is much smaller than the normal size, and should not be used. (In particular, while the NIFTI header indicates the correct dimensions, frames 83 onward are corrupted in that they contain a single intensity throughout the brain).

Completed Fix: Diffusion data for subject 103515 was completely reprocessed and the fixed version was available as part of the 900 Subjects Release.  The preprocessed dMRI for this subject from the 500 Subjects release should not be used (whether obtained via Connectome in a Box or downloaded online). The reprocessed files will be included in 900 Subjects Connectome in a Box versions and in the 900 Subjects update to the HCP Amazon S3 bucket.

A significant amount of head motion in the SBRef scan collected for the SOCIAL_LR task for S500 subject 748662 corrupted data for the entire scan. Therefore, the S500 released data for SOCIAL_LR of 748662 (including the unprocessed, minimally preprocessed, and level 2, cross run analyzed data) should not be used.   
We are in the process of removing the corrupt data from ConnectomeDB. SOCIAL tfMRI data for this subject downloaded from June 6, 2014-October 31, 2014 (or distributed in Connectome in a Box drives dated 20140805 or earlier) will contain the corrupted data and should not be used.

Completed Fix: The corrupt SOCIAL_LR tfMRI data for subject 748662 has been permanently removed from ConnectomeDB as of 10/31/2014. These files need to be removed from 500 Subjects Connectome in a Box dated 20140805 or earlier. These files have been removed in 500 Subjects Connectome in a Box versions dated 20141203 and later.

A problem with the early stages of the v3.1 (500 Subjects) preprocessing of diffusion data released for subject 173132 has been found. Diffusion data for this subject downloaded from June 6, 2014-October 31, 2014 (or distributed in Connectome in a Box drives dated 20140805 or earlier) should not be used.

Completed Fix: Diffusion data for subject 173132 has been completely reprocessed and the fixed version is available in ConnectomeDB as of 10/31/14.  These files will need to be replaced for 500 Subjects Connectome in a Box dated 20140805 or earlier. The reprocessed files are included in 500 Subjects Connectome in a Box versions dated 20141203 and later.

Minor Issues and Updates

Users who would like this script for current use can download it by clicking this link: prepare_level2_feat_analysis.sh 

Please also see the Advice for FEAT Analysis of HCP task fMRI data for more information on performing tfMRI analysis on HCP data.

Note: Users can also download task analysis data packages containing the results of individual subject, cross-run ("level 2") fMRI grayordinates analysis in FSL for each HCP task at various smoothing levels.  Analogous "level2" data packages for volume-based tfMRI processing are available at 4mm smoothing. Please see the HCP 500 Subjects Data Release Reference Manual for more details.

Q3 Release (September 23, 2013; more dMRI on November 26, 2013, all Q1-Q3 diffusion available January 31, 2014)

Major Issues

Two versions of the image reconstruction algorithm applied to dMRI and fMRI data have been used in HCP to date: version r177 for subjects scanned in Q1 through mid-Q3, version r227 for subjects scanned mid-Q3 and after.  We were able to retroactively recon all dMRI data using the newer r227 version. However, for the fMRI data, this was not possible and data reconstructed with both r177 and r227 versions remain in ConnectomeDB. The reconstruction version makes a notable signature on the data that can make a large difference in fMRI data analysis. Users wanting to use only data with the same reconstruction algorithm applied in their analyses can filter data on the ConnectomeDB dashboard by fMRI reconstruction version (in the Study Completion category). See Ramifications of Image Reconstruction Version Differences for more details.

Second-level QC on all subjects has revealed two Q2 subjects with structural brain abnormalities, specifically gray matter heterotopia. The abnormalities significantly affect functional and structural connectivity in these subjects, beyond what is considered normal variation. Therefore, we recommend that data from subjects 209733 and 528446 NOT be used in analyses.  

Subjects 209733 and 528446 were originally released in Q2 and have been available in ConnectomeDB and through Connectome in a Box since then. Both subjects were included in the Unrelated 40, Unrelated 80,  and Related 120 groups until the 500 subject release (June 6, 2014).

Completed Fix: All data from subjects 209733 and 528446 have been removed from ConnectomeDB as of the 500 subject release and will no longer be distributed by HCP.

The diffusion preprocessing output file containing the structural volume sampled at the same resolution as the diffusion data (necessary for analysis) was inadvertently left out of the dMRI data packages released for Q3:
e.g.  ${StudyFolder}/${SubjectID}/T1w/T1w_acpc_dc_restore_1.25.nii.gz
 
Completed Fix:
The dMRI data packages have been rebuilt with this file and as of 1/28/2014 are available via download from ConnectomeDB.

Users can also generate this file for previously downloaded or Connectome in a Box (CINAB) data, using the following FSL commands:

#Generate 1.25mm structural space for resampling the diffusion data into

${FSLDIR}/bin/flirt -interp spline -in T1w_acpc_dc_restore -ref T1w_acpc_dc_restore -applyisoxfm 1.25 -out T1w_acpc_dc_restore_1.25

${FSLDIR}/bin/applywarp --rel --interp=spline -i T1w_acpc_dc_restore -r T1w_acpc_dc_restore_1.25 -o T1w_acpc_dc_restore_1.25

Minor Issues and Updates

Users who would like this script for current use can download it by clicking this link: prepare_level2_feat_analysis.sh 

Please also see the Advice for FEAT Analysis of HCP task fMRI data for more information on performing tfMRI analysis on HCP data.

Note: With the Q3 and 500 Subjects release, users can also download task analysis data packages containing the results of individual subject, cross-run ("level 2") fMRI grayordinates analysis for each HCP task at various smoothing levels.  Analogous "level2" data packages for volume-based tfMRI processing at 4mm smoothing were made available as of March 2014 and also were part of the 500 Subjects released. Please see the HCP Q3 Data Release Reference Manual or the HCP 500 Subjects Data Release Reference Manual for more details.

Q2 Release (June 14, 2013)

including Q1 + Q2 v2 preprocessed data

Users who have the Q1+Q2 data: A zipped archive is available containing all of the fixed files for all Q1+Q2 subjects and a python script for patching existing download or Connectome in a box data, which places the fixed files in the HCP directory structure (details in the HCP Q2 Data Release Reference Manual and the associated directory structure Appendix III).   Download the Q1/Q2 data patch script here (requires ConnectomeDB login).

Major Issues

Second-level QC on all subjects has revealed two Q2 subjects with structural brain abnormalities, specifically gray matter heterotopia. The abnormalities significantly affect functional and structural connectivity in these subjects, beyond what is considered normal variation. Therefore, we recommend that data from subjects 209733 and 528446 NOT be used in analyses.  

Subjects 209733 and 528446 were originally released in Q2 and have been available in ConnectomeDB and through Connectome in a Box since then. Both subjects were included in the Unrelated 40, Unrelated 80,  and Related 120 groups until the 500 subject release (June 6, 2014).

Completed Fix:
All data from subjects 209733 and 528446 were removed from ConnectomeDB as of the 500 subject release and will no longer be distributed by HCP.

For one Q1+Q2 subject, DICOM to NIFTI failed to output correct number of volumes or lines in some tfMRI Relational data files:
 ${StudyFolder}/138231/MNINonLinear/Results/tfMRI_RELATIONAL_LR/*

Subject/scan affected:                                         file type                              Issue:                                                       

            138231/tfMRI_RELATIONAL_LR             4D volume                           volumes should be 232 instead of 106  

            138231/tfMRI_RELATIONAL_LR    all movement regressor files        lines should be 232 instead of 106

            138231/tfMRI_RELATIONAL_LR           Physio_log.txt                        triggers should be 232 instead of 105

Surface files for this subject will also be affected.

Completed Fix: DICOM2NIFTI was rerun and files are fixed as of the Q3 release (September 2013).

Users should either: 1) replace unprocessed and processed versions Q1+Q2 Relational data for subject 138231 by re-downloading the Relational packages (both unproc and preproc) for subject 138231, or 2) download and run the Q1/Q2 release patch script available here (requires ConnectomeDB login).

For some Q1+Q2 subjects, EV files for the Social tfMRI task were missing an entry for one or more blocks, and thus those EV files are inaccurate and incomplete. Specifically, there was a bug in the EV creation script used for the Q1 and Q2 releases for the SOCIAL task such that blocks were omitted from the EV file if the subject didn't respond during that block.

e.g.  ${StudyFolder}/${SubjectID}/MNINonLinear/Results/tfMRI_SOCIAL_RL/EVs/*.txt

Subjects/affected directories with this bug: Q1+Q2 Subjects with Shortened tfMRI Social EV *.txt Files

Completed Fix: These bugs have resolved for the Q3 release (September 2013).

Users should run the Q1/Q2 data patch script to correct all these incomplete files (download available here) (requires ConnectomeDB login).

Minor Issues

Completed fix: These bugs have resolved for the Q3 release (September 2013).

Users should run the Q1/Q2 data patch script to correct all these incomplete files (download available here) (requires ConnectomeDB login).

 There is a bug in the Q1+Q2 aparc.a2009s+aseg.nii.gz files (both in the T1w/ and MNINonLinear/ directories as shown below) for all subjects, such that all of cortex is labelled as 14175, which is wm_rh_S_temporal_transverse. The current files are corrupted in every subject and not usable for any analyses. 

 ${StudyFolder}/${SubjectID}/T1w/aparc.a2009s+aseg.nii.gz

 ${StudyFolder}/${SubjectID}/MNINonLinear/aparc.a2009s+aseg.nii.gz

Completed fix: These aparc files have been regenerated in Q1/Q2 subjects from pipeline intermediates and all Q1/Q2 packages have been updated with this change for Q3. The bug has been fixed in the PostFreeSurfer pipeline and Q3 subjects' data should not have this issue. 

Users should run the Q1/Q2 data patch script to correct all these incomplete files (download available here) (requires ConnectomeDB login).

In all minimally preprocessed fMRI data mapped to grayordinate surface vertices, e.g. 136833/MNINonLinear/Results/rfMRI_REST2_LR/rfMRI_REST2_LR_Atlas.dtseries.nii, there are some empty vertices (timeseries = 0s) very near the medial wall. This is because the FreeSurfer defined medial wall sometimes extends outside the atlas medial wall used to define the grayordinates space.  Due to a bug in the PostFreeSurfer Pipeline, these vertices were set to zero because the affected subject's specific medial wall ROI was larger than the atlas medial wall ROI. This bug appears in approximately half the subjects' right and/or left hemispheres (too many to list here).

This bug should not create a problem for most users because the empty vertices should always be located around the medial wall (where the data is not very good anyway). However, the zero values could cause some algorithms to crash if they can't cope with zeros in CIFTI/surface data.The tfMRI analysis pipeline is not affected by this bug and the issue is fixed in the analyzed task data.  FIX is not negatively affected by this, but the bug remains in the fix cleaned timeseries.

Planned Fix: This bug is not trivial to fix for all Q1, Q2, and Q3 subjects without running all these subjects through a new version of the minimally preprocessing pipelines (specifically the PostFreeSurfer pipeline, see Glasser et al. 2013 for details), which will take a substantial amount of computing time. To date (August 2013), we have not decided exactly when we will do this.

Users who need to fix the files before we make the permanent fix, a relatively simple DIY fix of affected files is available using the command "-cifti-dilate" in wb_command.

Missing Files

Near Completed fix: For all but the last two subjects, 131924 and 677968, the data has been recovered and will be released as part of the Q3 release in the ConnectomeDB. For the two remaining subjects, we have sent paper forms for them to refill their responses to the questions asked at their subject visit, which we will add to ConnectomeDB once they are received.

Users who have the Q1+Q2 data should re-download the fixed behavioral data for all subjects available here or on the ConnectomeDB splash page.