A primer for the upcoming developing Human Connectome Project data release; presented at the Big Data Little Brains conference in Chapel Hill, May 2018
1. The Developing Human Connectome Project (dHCP):
Towards a dynamic map of evolving brain
connectivity, reflecting fetal and early neonatal
periods
Dr Emma C. Robinson
@emrobSci
emma.robinson@kcl.ac.uk
https://emmarobinson01.com/
2. The Developing Human Connectome Project
• Mapping the emergence of brain
connectivity from 20-44 weeks PMA
• ~1500 scans
• Acquisitions (MRI):
Resting state fMRI
Multi-shell HARDI
Structural T1 and T2
• Supported by
Genetic samples
Cognitive test scores/eye tracking
Demographics
http://www.developingconnectome.org/
c/o Dr Bernard Kainz, Imperial College
3. The Developing Human Connectome Project
• Mapping the emergence of brain
connectivity from 20-44 weeks PMA
• ~1500 scans
• Acquisitions (MRI):
Resting state fMRI
Multi-shell HARDI
Structural T1 and T2
• Supported by
Genetic samples
Cognitive test scores/eye tracking
Demographics
http://www.developingconnectome.org/
c/o Dr Bernard Kainz, Imperial College
4. The Developing Human Connectome Project
• Mapping the emergence of brain
connectivity from 20-44 weeks PMA
• ~1500 scans
• Acquisitions (MRI):
Resting state fMRI
Multi-shell HARDI
Structural T1 and T2
• Supported by
Genetic samples
Cognitive test scores/eye tracking
Demographics
http://www.developingconnectome.org/
c/o Dr Bernard Kainz, Imperial College
7. Challenges of working
with developing data
• Developing data is affected by
Motion (severe cases account for < 2% )
8. Challenges of working
with developing data
• Developing data is affected by
Motion (severe cases account for < 2% )
Limited scan times
9. Challenges of working
with developing data
• Developing data is affected by
Motion (severe cases account for < 2% )
Limited scan times
Relatively low resolution
10. Challenges of working
with developing data
• Developing data is affected by
Motion (severe cases account for < 2% )
Limited scan times
Relatively low resolution
Inverted contrast
11. Challenges of working
with developing data
• Developing data is affected by
Motion (severe cases account for < 2% )
Limited scan times
Relatively low resolution
Inverted contrast
spatio-temporal evolution
14. Neonatal Structural Pipeline
• Reconstruction with motion
correction
• Turbo Spin Echo (TSE) T2
• Two stacks of 2D slices
• 0.8x0.8x1.6 mm
(image courtesy: M. Fogtmann, IEEE TMI 2014)
15. Neonatal Structural Pipeline
• Reconstruction with motion
correction
• Turbo Spin Echo (TSE) T2
• Two stacks of 2D slices
• 0.8x0.8x1.6 mm
• Slice to volume -> 0.5mm3
Cordero-Grande, Lucilio, et al. "Three-dimensional
motion corrected sensitivity encoding reconstruction
for multi-shot multi-slice MRI: Application to
neonatal brain imaging." MRM (2018)”
16. Neonatal Structural Pipeline
Makropoulos and Robinson et al. The Developing Human Connectome Project: a Minimal Processing Pipeline for Neonatal Cortical Surface
Reconstruction. NeuroImage (2018)
• Reconstruction with motion
correction
17. Neonatal Structural Pipeline
Makropoulos and Robinson et al. The Developing Human Connectome Project: a Minimal Processing Pipeline for Neonatal Cortical Surface
Reconstruction. NeuroImage (2018)
• Reconstruction with motion
correction
• Tissue segmentation
High intensity white matter correction
Makropoulos, Antonios, et al. "Automatic whole
brain MRI segmentation of the developing neonatal
brain." IEEE transactions on medical imaging 33.9
(2014): 1818-1831.
18. Neonatal Structural Pipeline
Brain extract
Bias correct T1T2
Align
White
PialMid-thickness Inflated Very Inflated Sphere
T1/T2w ratio
A. Pre-Processing
F
G
C
H
B
DE
I
I
Myelin MapSulcal DepthCurvatureThickness
Segmentation
Makropoulos and Robinson et al. The Developing Human Connectome Project: a Minimal Processing Pipeline for Neonatal Cortical Surface
Reconstruction. NeuroImage (2018)
• Reconstruction with motion
correction
• Tissue segmentation
• Surface mesh modelling
19. Neonatal Structural Pipeline
Brain extract
Bias correct T1T2
Align
White
PialMid-thickness Inflated Very Inflated Sphere
T1/T2w ratio
A. Pre-Processing
F
G
C
H
B
DE
I
I
Myelin MapSulcal DepthCurvatureThickness
Segmentation
Makropoulos and Robinson et al. The Developing Human Connectome Project: a Minimal Processing Pipeline for Neonatal Cortical Surface
Reconstruction. NeuroImage (2018)
• Reconstruction with motion
correction
• Tissue segmentation
• Surface mesh modelling
• Feature Extraction
20. Neonatal Surface QC
2 raters rated
• 43 images
• Patches of size
50x50x50mm
• White surface only
Comparison of intensity-based surface refinement (green) to segmentation result (yellow)
Example QC from single rater
21. Neonatal fMRI
Pipeline
Fitzgibbon, Sean P., et al. "The developing Human Connectome Project (dHCP): minimal functional
pre-processing pipeline for neonates." Fifth Biennial Conference on Resting State and Brain
Connectivity. 2016.
22. Neonatal fMRI
Pipeline
• FIELDMAP
Fitzgibbon, Sean P., et al. "The developing Human Connectome Project (dHCP): minimal functional
pre-processing pipeline for neonates." Fifth Biennial Conference on Resting State and Brain
Connectivity. 2016.
23. Neonatal fMRI
Pipeline
• FIELDMAP
• Motion & Distortion
Correction (MCDC)
Fitzgibbon, Sean P., et al. "The developing Human Connectome Project (dHCP): minimal functional
pre-processing pipeline for neonates." Fifth Biennial Conference on Resting State and Brain
Connectivity. 2016.
24. Neonatal fMRI
Pipeline
• FIELDMAP
• Motion & Distortion
Correction (MCDC)
• REGISTRATION
Fitzgibbon, Sean P., et al. "The developing Human Connectome Project (dHCP): minimal functional
pre-processing pipeline for neonates." Fifth Biennial Conference on Resting State and Brain
Connectivity. 2016.
25. Neonatal fMRI
Pipeline
• FIELDMAP
• Motion & Distortion
Correction (MCDC)
• REGISTRATION
• ICA+FIX
Fitzgibbon, Sean P., et al. "The developing Human Connectome Project (dHCP): minimal functional
pre-processing pipeline for neonates." Fifth Biennial Conference on Resting State and Brain
Connectivity. 2016.
26. Neonatal fMRI
Pipeline
• FIELDMAP
• Motion & Distortion
Correction (MCDC)
• REGISTRATION
• ICA+FIX
• NUISANCE REGRESSION
Fitzgibbon, Sean P., et al. "The developing Human Connectome Project (dHCP): minimal functional
pre-processing pipeline for neonates." Fifth Biennial Conference on Resting State and Brain
Connectivity. 2016.
27. Neonatal fMRI
Pipeline
• FIELDMAP
• Motion & Distortion
Correction (MCDC)
• REGISTRATION
• ICA+FIX
• NUISANCE REGRESSION
• SAMPLE TO SURFACE
Fitzgibbon, Sean P., et al. "The developing Human Connectome Project (dHCP): minimal functional
pre-processing pipeline for neonates." Fifth Biennial Conference on Resting State and Brain
Connectivity. 2016.
28. Neonatal fMRI
Pipeline
• FIELDMAP
• Motion & Distortion
Correction (MCDC)
• REGISTRATION
• ICA+FIX
• NUISANCE REGRESSION
• SAMPLE TO SURFACE
• QC
Fitzgibbon, Sean P., et al. "The developing Human Connectome Project (dHCP): minimal functional
pre-processing pipeline for neonates." Fifth Biennial Conference on Resting State and Brain
Connectivity. 2016.
31. Neonatal dMRI Pipeline
• 300 diffusion volumes (20 b0)
• b = 400 s/mm2 (64)
• b = 1000 s/mm2 (88)
• b = 2600 s/mm2 (128)
32. Neonatal dMRI Pipeline
• 300 diffusion volumes (20 b0)
• b = 400 s/mm2 (64)
• b = 1000 s/mm2 (88)
• b = 2600 s/mm2 (128)
• Correction for eddy currents, susceptibility and motion
performed with FSL’s Eddy
Jesper L. R. et al. An integrated approach to correction for off-resonance effects and
subject movement in diffusion MR imaging. NeuroImage, 125:1063-1078, 2016.
33. Neonatal dMRI Pipeline
• 300 diffusion volumes (20 b0)
• b = 400 s/mm2 (64)
• b = 1000 s/mm2 (88)
• b = 2600 s/mm2 (128)
• Correction for eddy currents, susceptibility and motion
performed with FSL’s Eddy
• Virtual dissection (atlas based)
Bastiani et al., Automated
processing pipeline for
neonatal diffusion MRI in the
developing Human
Connectome Project.
NeuroImage (under review).
34. • Micro-structural parameter estimates using NODDI (Zhang et al 2012)
Microstructure Tracts (virtual dissection)
Neonatal dMRI Pipeline
Bastiani et al., Automated processing
pipeline for neonatal diffusion MRI in the
developing Human Connectome Project.
NeuroImage (under review).
38 39 40 4138 39 40 41
35. dHCP spatio-temporal atlases
• New volumetric and surface templates spanning 36-44 weeks
gestation
Andreas Schuh et al.
Unbiased construction
of a temporally
consistent
morphological atlas of
neonatal brain
development
(under review)
36. dHCP spatio-temporal atlases
• New volumetric and surface templates spanning 36-44 weeks
gestation
Jelena Bozek et al.
Construction of a Neonatal
Cortical Surface Atlas Using
Multimodal Surface
Matching in the Developing
Human Connectome Project
(under revision)
37. dHCP spatio-temporal atlases
• New volumetric and surface templates spanning 36-44 weeks
gestation
Jelena Bozek et al.
Construction of a Neonatal
Cortical Surface Atlas Using
Multimodal Surface
Matching in the Developing
Human Connectome Project
(under revision)
38. dHCP spatio-temporal atlases
• New volumetric and surface templates spanning 36-44 weeks
gestation
Jelena Bozek et al.
Construction of a Neonatal
Cortical Surface Atlas Using
Multimodal Surface
Matching in the Developing
Human Connectome Project
(under revision)
39. Surface-based alignment of neonatal
cortical surfaces
• Spherical framework for cortical surface registration: MSM
• Use low resolution control point grids to constrain the deformation
• Optimised using discrete methods
Robinson, Emma C., et al. "MSM: a new flexible framework for Multimodal Surface Matching." Neuroimage (2014)
40. Surface-based alignment of neonatal
cortical surfaces
• Spherical framework for cortical surface registration: MSM
• Use low resolution control point grids to constrain the deformation
• Optimised using discrete methods
Robinson, Emma C., et al. "MSM: a new flexible framework for Multimodal Surface Matching." Neuroimage (2014)
41. Surface-based alignment of neonatal
cortical surfaces
• Spherical framework for cortical surface registration: MSM
• Use low resolution control point grids to constrain the deformation
• Optimised using discrete methods
data cost: i.e. correlation, MNI, SSD
Regularisation cost to encourage smoother warp
Robinson, Emma C., et al. "MSM: a new flexible framework for Multimodal Surface Matching." Neuroimage (2014)
42. SAS SSS
a)
e)
b)
x LABEL POINT
OPTIMAL LABEL
CONTROL POINT
c)
f)
d)
MSS
SSS + G
TSS
DAS TAS
g)
F
Finding trends in longitudinal cortical
development
MSM now also allows
smooth deformation
of cortical anatomies
Robinson, Emma C., et al.
"Multimodal surface matching
with higher-order smoothness
constraints." NeuroImage
(2018).
43. New MSM: finding trends in
longitudinal cortical development
• 24 very preterm infants (born <30 weeks PMA, 15 male, 15 female)
• scanned 2-4 times before or at term-equivalent (36-40 weeks PMA)
Garcia, Kara E., Robinson E.C. et al. "Dynamic patterns of cortical expansion during folding of the preterm human brain." PNAS (2018)
44. Deep Learning for Brain Imaging
Transfer learning: Applied to motion artefact
detection of neonatal diffusion MRI (dMRI)
Data from
Developing Human
Connectome
Project (dHCP) Kelly, C., Pietsch, M., Counsell, S., & Tournier, J. D. Transfer learning
and convolutional neural net fusion for motion artefact detection.
45. Deep Learning for Brain Imaging
Transfer learning: Applied to motion artefact
detection of neonatal diffusion MRI (dMRI)
•Like fMRI highly
sensitive to motion
Data from
Developing Human
Connectome
Project (dHCP) Kelly, C., Pietsch, M., Counsell, S., & Tournier, J. D. Transfer learning
and convolutional neural net fusion for motion artefact detection.
Red boxed highlight motion artifacted slices
46. Deep Learning for Brain Imaging
Transfer learning: Applied to motion artefact
detection of neonatal diffusion MRI (dMRI)
•Like fMRI highly
sensitive to motion
•Standard practice to
remove noisy slices
Data from
Developing Human
Connectome
Project (dHCP) Kelly, C., Pietsch, M., Counsell, S., & Tournier, J. D. Transfer learning
and convolutional neural net fusion for motion artefact detection.
Red boxed highlight motion artifacted slices
47. Deep Learning for Brain Imaging
Transfer learning: Applied to motion artefact
detection of neonatal diffusion MRI (dMRI)
•Like fMRI highly
sensitive to motion
•Standard practice to
remove noisy slices
•Train CNN classifier
using transfer learning
Data from
Developing Human
Connectome
Project (dHCP) Kelly, C., Pietsch, M., Counsell, S., & Tournier, J. D. Transfer learning
and convolutional neural net fusion for motion artefact detection.
Red boxed highlight motion artifacted slices
48. Deep Learning for Medical Imaging
Transfer learning: Applied to motion artefact
detection of neonatal diffusion MRI (dMRI)
Kelly, C., Pietsch, M., Counsell, S., & Tournier, J. D. Transfer learning and convolutional neural net fusion for motion artefact detection.
49. Deep Learning for Medical Imaging
Transfer learning: Applied to motion artefact
detection of neonatal diffusion MRI (dMRI)
•Trained on 36 subjects
Kelly, C., Pietsch, M., Counsell, S., & Tournier, J. D. Transfer learning and convolutional neural net fusion for motion artefact detection.
50. Deep Learning for Medical Imaging
Transfer learning: Applied to motion artefact
detection of neonatal diffusion MRI (dMRI)
•Trained on 36 subjects
•Multiple CNNs trained
on different categories
of dMRI data
Kelly, C., Pietsch, M., Counsell, S., & Tournier, J. D. Transfer learning and convolutional neural net fusion for motion artefact detection.
51. Deep Learning for Medical Imaging
Transfer learning: Applied to motion artefact
detection of neonatal diffusion MRI (dMRI)
•Trained on 36 subjects
•Multiple CNNs trained
on different categories
of dMRI data
•Output of predictions
merged by random
forest
94.8%-99.8%
accuracy
Human level
~99.25%
Kelly, C., Pietsch, M., Counsell, S., & Tournier, J. D. Transfer learning and convolutional neural net fusion for motion artefact detection.
52. Fetal Pipeline
!18
• T1/T2
• 0.75 mm isotropic
• 141 Diffusion gradients
• b = 0 s/mm2 (15)
• b = 400 s/mm2 (46)
• b = 1000 s/mm2 (80)
• 2mm isotropic
• fMRI
• 12mins 35 secs
• 2.2mm isotropic
Example T2
53. Fetal Pipeline
!18
• T1/T2
• 0.75 mm isotropic
• 141 Diffusion gradients
• b = 0 s/mm2 (15)
• b = 400 s/mm2 (46)
• b = 1000 s/mm2 (80)
• 2mm isotropic
• fMRI
• 12mins 35 secs
• 2.2mm isotropic
Example T2
54. Fetal Pipeline
!18
• T1/T2
• 0.75 mm isotropic
• 141 Diffusion gradients
• b = 0 s/mm2 (15)
• b = 400 s/mm2 (46)
• b = 1000 s/mm2 (80)
• 2mm isotropic
• fMRI
• 12mins 35 secs
• 2.2mm isotropic
Example T2
55. • SLICE-WISE MOTION
CORRECTION AND DYNAMIC
DISTORTION CORRECTION
• SEGMENTATION /
REGISTRATION
• ICA+FIX
• SAMPLE TO SURFACE
• NUISANCE REGRESSION
• QC
Fetal Pipeline: fMRI Piloting
!19
Neonatal
22 - 37 week gestational age scans
Group Mean SNR
60. Fetal Pipeline: dMRI Piloting
!21
• Slice to Volume
Reconstruction (with
motion correction)
• “Multi-shell SHARD reconstruction
from scattered slice diffusion MRI
data in the neonatal brain.” Daan
Christiaens et al ISMRM 2018
• Deprez, Maria, et al. "Higher order
spherical harmonics reconstruction
of fetal diffusion MRI with intensity
correction." bioRxiv (2018): 297341.
INPUT
SHARD RECONSTRUCTION
61. Fetal Pipeline: dMRI Piloting
!21
• Slice to Volume
Reconstruction (with
motion correction)
• “Multi-shell SHARD reconstruction
from scattered slice diffusion MRI
data in the neonatal brain.” Daan
Christiaens et al ISMRM 2018
• Deprez, Maria, et al. "Higher order
spherical harmonics reconstruction
of fetal diffusion MRI with intensity
correction." bioRxiv (2018): 297341.
• Spherical
Deconvolution fit
• Constrained
• b 1000
62. Data Releases
!22
•1st Pilot data release
• https://data.developingconnectome.org/app/template/
Login.vm
• 40 neonatal subjects:
• T1, T2, fMRI and dMRI volumes (minimally processed)
• output of surface extraction pipelines
•2nd Major data release
• Expected summer 2018
• For queries on data releases and pipelines see https://
neurostars.org/tags/developing-hcp
63. Data Releases
!23
•dHCP structural pipeline
• https://github.com/BioMedIA/dhcp-structural-
pipeline
• Includes docker installation
• Contact j.cupitt@imperial.ac.uk
64. Acknowledgements
!24
• Professor A. David Edwards (PI)
• Professor Jo Hajnal (PI)
• Dr Lucillio Cordero Grande
• Dr Anthony Price
• Dr Maria Deprez
• Dr Chris Kelly
• Max Pietsch
• Daan Christiaens
• Dr Donald Tournier
• Dr Emer Hughes
http://www.developingconnectome.org/teams-and-collaborators-v2/
• Professor Daniel Rueckert (PI)
• Dr Antonios Makropoulos
• Dr Andreas Schuh
• Dr Jonathan Palmbach-Passerat
• Dr John Cupitt
• Dr Jianling Gao
• Professor Steve Smith (PI)
• Professor Mark Jenkinson
• Dr Eugene Duff
• Dr Matteo Bastiani
• Dr Sean Fitzgibbon
• Dr Saad Jbabdi
• Dr Stam Sotiropoulos
• Dr Jelena Bozek