dHCP Fourth Data Release (April 2024)

The structural pipeline

The structural pipeline was encapsulated as a Docker container image and run via the OpenMOLE¹⁰ platform on a local cluster.

Inputs and outputs

From reconstruction pipeline: rawdata/sub-{subid}/ses-{sesid}

The structural pipline generates the files listed in the Structural pipeline section of the directory structure summary.

Operation

1. Registration

2. Segmentation

   1. Structural scans are pre-processed by first running bias correction using the N4 algorithm¹.

   2. Scans are then brain extracted using BET² from FSL.

   3. Segmentation of the T2w volume is performed using the DRAW-EM algorithm³. DRAW-EM is an atlas-based segmentation technique that segments the volumes into 87 regions (see region names). Manually labelled atlases, annotated by an expert neuroanatomist⁴, are registered to the volume and their labels are fused to the subject space to provide structure priors. Segmentation is then performed with an Expectation-Maximization scheme that combines the structure priors and an intensity model of the volume. The 87 regions are further merged to provide the tissue segmentation (see tissue types).

   4. All T1 weighted images have been pre-aligned to the T2w volumes using rigid alignment.

   5. Both T1w and T2w volumes are defaced for anonymization based on registration and transformation of a manually annotated face mask.

3. Surface extraction

   1. Surface mesh extraction is performed with the method described elsewhere⁵. A white matter mask enclosing the white surface is computed by merging the white matter and the subcortical structures with the exception of the brainstem and the cerebellum. Similarly, a pial mask is computed by merging the grey matter structure with the white matter mask. The white and pial surfaces of the left and right hemispheres are then reconstructed with the method outlined in 5 using a deformable model. The model in 5 includes forces to avoid self-intersections and includes an image-based refinement step that corrects regions such as deep sulci mislabelled by the volumetric segmentation.

   2. Midthickness surfaces are generated as the middle surface between the white and pial surfaces. The midthickness surface is computed using the Euclidean distance between corresponding points of the white and pial surface.

   3. Spherical projection is performed6, and it is based on the inflated white matter surface. The inflated white matter surface is produced in a similar manner as in the FreeSurfer pipeline⁷. Inflated and very inflated surfaces used for visualisation are generated similarly⁸.

   4. The following metrics are further estimated from the surfaces: curvature, thickness, sulcal depth, T1w/T2w myelin, labels (projected from the volume). All surfaces have one-to-one vertex correspondence for all points on the surface ensuring that the same vertex indexes the same point, in the same relative position, on the anatomy for all surfaces. Please note that due to the relatively large voxels (causing partial volume) and/or the uneven vertex sampling for gyri (relative to sulci) we observe some artifacts in the myelin and thickness metric files, which resemble the folding patterns. For this, we offer a ‘corrected’ version of thickness maps (corr_thickness); however, this results only from a linear regressesion based correction. We advise careful interpretation of these maps and have chosen at this time not to do the same for myelin.

4. Surface registration

   1. A new symmetric and extended version of the neonatal surface template¹¹ is available from the brain-development.org website

   2. All surfaces have been nonlinearly aligned to the template using cortical folding-driven aligment (implemented with MSM^11,12); alignment has been optimised with relatively weak regularisation to push cortical correspondence of folds in the frontal lobe. At this time we have no evidence of any negative impact on the dMRI and fMRI correspondence through doing this. However, scripts to re-run registration wih modifed parameters are available from here

   3. We release only the registration warp file (sphere.reg.surf.gii). To obtain the warped metric files and surfaces in template space please run the resampling scripts

The structural pipeline is described in detail elsewhere⁹.

Quality Control/Assurance

The dHCP structural data set contains MR imaging of whole brain anatomy across a wide gestational age. The data collected contains rapidly changing anatomical variation across age, that inherently provides challenges for automatic segmentation processes. For this reason, we carried out a visual inspection of the segmentation pipeline on a random set of 30 datasets (gestational age range at scan 27.14 weeks-44.14 weeks).

The following anatomical segmentations were assessed:

• Cortical ribbon
• Ventricular space including the cavum septum pellucidum
• Whole brain white matter
• Deep grey matter (basal ganglia)

The results showed some small areas of common segmentation errors across the dataset that can be easily correctable with some fine editing:

1. The cortical ribbon segmentation can contain small regions of highly convoluted cortex that may not be sufficiently delineated due to partial volume effects and there is consistent mis-registration of the cortex in the medial temporal lobe.
2. The anterior cavum septum pellucidum can be mis-labelled as ventricular space.
3. The basal ganglia/deep grey matter anatomy may have a small chunk of anatomy mislabelled as white matter.
4. The whole brain white matter segmentation was found to be consistently accurate across subjects.

We therefore recommend that visual inspection of the segmentation data in this cohort is carried out before inclusion in analysis.

References

1. Tustison, N. J., Avants, B. B., Cook, P. A., Zheng, Y., Egan, A., Yushkevich, P. A. and Gee, J. C. N4ITK: improved N3 bias correction. IEEE transactions on medical imaging (2010) 29(6): 1310-1320. DOI: 10.1109/TMI.2010.2046908

2. Smith, S. M. Fast robust automated brain extraction Human Brain Mapping (2002), 17(3): 143-55. DOI: 10.1002/hbm.10062

3. Makropoulos, A., Gousias I. S., Ledig C., Aljabar P., Serag A, Hajnal J. V., Edwards A. D., Counsell S. J., and Rueckert D. Automatic whole brain MRI segmentation of the developing neonatal brain IEEE transactions on medical imaging (2014), 33(9): 1818-1831. DOI: 10.1109/TMI.2014.2322280

4. Gousias, I. S., Edwards A. D., Rutherford M. A., Counsell S. J., Hajnal J. V., Rueckert D., and Hammers, A. Magnetic resonance imaging of the newborn brain: manual segmentation of labelled atlases in term-born and preterm infants Neuroimage (2012), 62(3): 1499-1509. DOI: 10.1016/j.neuroimage.2012.05.083

5. Schuh, A., Makropoulos, A., Wright, R., Robinson, E. C., Tusor, N., Steinweg, J., Hughes, E., Cordero Grande, L., Price, A., Hutter, J., Hajnal, J., and Rueckert, D. A deformable model for reconstruction of the neonatal cortex IEEE 14th International Symposium on Biomedical Imaging 2017. DOI: 10.1109/ISBI.2017.7950639

6. Elad, A., Keller, Y., and Kimmel, R. Texture Mapping via Spherical Multidimensional Scaling International Conference on Scale-Space Theories in Computer Vision (2005): 443–455.

7. Fischl, B., Sereno M. I., and Dale, A. M. Cortical surface-based analysis: II: inflation, flattening, and a surface-based coordinate system Neuroimage (1999), 9(2): 195-207. DOI: 10.1006/nimg.1998.0396

8. Glasser, M., Sotiropoulo, S. N., Wilson, J. A, Coalson, T. S, Fischl, B., Andersson, L., Xu, J., Jbabdi, S., Webster, M., Polimeni, J. R., Van Essen, D. C., Jenkinson, M., WU-Minn HCP Consortium. The minimal preprocessing pipelines for the Human Connectome Project NeuroImage (2013)., 80: 105–124. DOI: 10.1016/j.neuroimage.2013.04.127

9. Makropoulos, A., Robinson, E.C., Schuh, A., Wright, R., Fitzgibbon, S., Bozek, J., Counsell, S. J., Steinweg, J., Vecchiato, K., Passerat-Palmbach, J., Lenz, G., Mortari, F., Tenev, T., Duff, E. P., Bastiani, M., Cordero-Grande, L., Hughes, E., Tusor, N., Tournier, J. D., Hutter, J., Price, A. N., Teixeira, R. P. A. G., Murgasova M, Victor, S., Kelly, C., Rutherford, M. A., Smith, S. M., Edwards, A. D., Hajnal, J. V., Jenkinson, M., and Rueckert, D. The developing Human Connectome Project: a Minimal Processing Pipeline for Neonatal Cortical Surface Reconstruction NeuroImage (2018), 173: 88-112. DOI: 10.1016/j.neuroimage.2018.01.054

10. Passerat-Palmbach, J., Reuillon, R., Leclaire, M., Makropoulos, A., Robinson, E.C., Parisot, S., and Rueckert, D. Reproducible Large-Scale Neuroimaging Studies with the OpenMOLE Workflow Management System Frontiers in Neuroinformatics (2017), 11. DOI: 10.3389/fninf.2017.00021

11. Bozek, J., Makropoulos, A., Schuh, A., Fitzgibbon, S., Wright, R., Glasser, M.F., Coalson, T.S., O’Muircheartaigh, J., Hutter, J., Price, A.N., Cordero-Grande, L. Teixeira, R. P. A. G., Hughes, E., Tusor, N., Pegoretti Baruteau, K., Rutherford, M. A., Edwards, A. D., Hajnal, J. V., Smith, S. M., Rueckert, D., Jenkinson, M., Robinson, E.C, Construction of a neonatal cortical surface atlas using Multimodal Surface Matching in the Developing Human Connectome Project. NeuroImage (2018) 179: 11-29. DOI: 10.1016/j.neuroimage.2018.06.018

12. Robinson, E.C., Jbabdi, S., Glasser, M.F., Andersson, J., Burgess, G.C., Harms, M.P., Smith, S.M., Van Essen, D.C. and Jenkinson, M. MSM: a new flexible framework for Multimodal Surface Matching. Neuroimage (2014), 100: 414-26. DOI: 10.1016/j.neuroimage.2014.05.069

13. Robinson, E.C., Garcia, K., Glasser, M.F., Chen, Z., Coalson, T.S., Makropoulos, A., Bozek, J., Wright, R., Schuh, A., Webster, M. and Hutter, J., Price, A.N., Cordero-Grande, L. Hughes, E., Tusor, N., Bayley, P.V., Van Essen, D.C., Smith, S. M., Edwards, A. D., Hajnal, J. Jenkinson, M., Glocker, B., Rueckert, D., Multimodal surface matching with higher-order smoothness constraints. Neuroimage (2018), 167: 453-465. DOI: 10.1016/j.neuroimage.2017.10.037