![Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data
“Spine Generic Protocol” initiative
5Alley et al., ISMRM 2018 ; “white paper” in preparation
• Solution: Establish a consensus set of qMRI acquisition
parameters for the spinal cord at 3T. [Alley18]
• Similar to NINDS-CDE initiative, but focusing on qMRI
metrics and not specific to particular diseases
• 21 international sites involved in the optimization
• Protocol available for GE, Philips and Siemens at:
www.spinalcordmri.org/protocols](https://image.slidesharecdn.com/20190122sc-workshop-190122004341/85/20190122_cohenadad_sc-mri-workshop-5-320.jpg)
![Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data
• Solution: Establish a consensus set of qMRI acquisition
parameters for the spinal cord at 3T. [Alley18]
• Similar to NINDS-CDE initiative, but focusing on qMRI
metrics and not specific to particular diseases
• 21 international sites involved in the optimization
• Protocol available for GE, Philips and Siemens at:
www.spinalcordmri.org/protocols
• Standard Operating Procedure (SOP)
• Intuitive procedure, with pictures
• Document available online
“Spine Generic Protocol” initiative
6Alley et al., ISMRM 2018](https://image.slidesharecdn.com/20190122sc-workshop-190122004341/85/20190122_cohenadad_sc-mri-workshop-6-320.jpg)
![Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data
• Solution: Establish a consensus set of qMRI acquisition
parameters for the spinal cord at 3T. [Alley18]
• Similar to NINDS-CDE initiative, but focusing on qMRI
metrics and not specific to particular diseases
• 21 international sites involved in the optimization
• Protocol available for GE, Philips and Siemens at:
www.spinalcordmri.org/protocols
• Standard Operating Procedure (SOP)
• Intuitive procedure, with pictures
• Document available online
• Publicly-available dataset:
• Multi-site, single subject (n=18)
• Multi-site, multi-subject (n~200)
“Spine Generic Protocol” initiative
7Alley et al., ISMRM 2018](https://image.slidesharecdn.com/20190122sc-workshop-190122004341/85/20190122_cohenadad_sc-mri-workshop-7-320.jpg)
![Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data
Multi-site, multi-subjects
• Processing of 32 subjects done in ~1h30 (2.2GHz x 40, 512 RAM)
• Example of QC: https://osf.io/z6stq/
• Preliminary results of cord CSA averaged between C2-C3 vertebral levels:
14Sites: Barcelona (Siemens-prisma_fit, n=6), Milan (Siemens-prisma, n=6), Sherbrooke (Philips-Ingenia, n=7), UCL (Philips-Ingenia, n=6), UNF (Siemens-prisma_fit, n=6)
0
15
30
45
60
75
90
2019-01-21 12:01:50 2019-01-21 12:03:28 2019-01-21 12:09:32
CSA[mm2]
T2w
Subjects
0
15
30
45
60
75
90
2019-01-21 11:57:39 2019-01-21 11:59:06 2019-01-21 12:05:15
CSA[mm2]
Subjects
T1w](https://image.slidesharecdn.com/20190122sc-workshop-190122004341/85/20190122_cohenadad_sc-mri-workshop-14-320.jpg)
20190122_cohenadad_sc-mri-workshop
- 1. Standardizing acquisition and processing of spinal cord MRI data Spinal Cord MRI Workshop January 22nd, 2019, London, UK Julien Cohen-Adad, PhD Associate Professor, Ecole Polytechnique de Montreal Associate Director, Functional Neuroimaging Unit, University of Montreal Canada Research Chair in Quantitative Magnetic Resonance Imaging
- 2. Outline 1. Standardize spinal cord MRI acquisition 2. Standardize spinal cord MRI processing Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data 2
- 3. Outline 1. Standardize spinal cord MRI acquisition 2. Standardize spinal cord MRI processing Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data 3
- 4. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data qMRI: What protocol to use? 4 If a researcher not familiar with qMRI was to start a project involving spinal cord imaging, what sequence and parameters to choose? qMTMTR MT_sat MWF CHARMED NODDI DTI DKI g-ratio ultra-short TE T1 T2* T2 PD This can become overwhelming!
- 5. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data “Spine Generic Protocol” initiative 5Alley et al., ISMRM 2018 ; “white paper” in preparation • Solution: Establish a consensus set of qMRI acquisition parameters for the spinal cord at 3T. [Alley18] • Similar to NINDS-CDE initiative, but focusing on qMRI metrics and not specific to particular diseases • 21 international sites involved in the optimization • Protocol available for GE, Philips and Siemens at: www.spinalcordmri.org/protocols
- 6. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data • Solution: Establish a consensus set of qMRI acquisition parameters for the spinal cord at 3T. [Alley18] • Similar to NINDS-CDE initiative, but focusing on qMRI metrics and not specific to particular diseases • 21 international sites involved in the optimization • Protocol available for GE, Philips and Siemens at: www.spinalcordmri.org/protocols • Standard Operating Procedure (SOP) • Intuitive procedure, with pictures • Document available online “Spine Generic Protocol” initiative 6Alley et al., ISMRM 2018
- 7. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data • Solution: Establish a consensus set of qMRI acquisition parameters for the spinal cord at 3T. [Alley18] • Similar to NINDS-CDE initiative, but focusing on qMRI metrics and not specific to particular diseases • 21 international sites involved in the optimization • Protocol available for GE, Philips and Siemens at: www.spinalcordmri.org/protocols • Standard Operating Procedure (SOP) • Intuitive procedure, with pictures • Document available online • Publicly-available dataset: • Multi-site, single subject (n=18) • Multi-site, multi-subject (n~200) “Spine Generic Protocol” initiative 7Alley et al., ISMRM 2018
- 8. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data “Spine Generic Protocol” initiative 8Alley et al., ISMRM 2018 Total acquisition time: 20-30 minutes Diffusion Tensor Imaging - Demyelination in WM - Axon degeneration Magnetization Transfer - Demyelination in WM 2D multi-echo GRE - Gray matter atrophy 3D T1w 1mm - Brain & Spine assessment 3D T2w 0.8mm - Cord atrophy
- 9. Results in one “cooperative” subject scanned at 18 sites
- 10. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data T1w: Cord CSA (single subject) 10Alley et al., ISMRM 2018 Strongfiltering Cord cross-sectional area (CSA) averaged between C2-C3 A P R L 1 x 1 x 1 mm3 GE Philips Siemens x 71.04 ± 0.69 75.80 ± 0.71 74.12 ± 1.85
- 11. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data MTR (single subject) 11Alley et al., ISMRM 2018 A P R L 0.9 x 0.9 x 5 mm3 MTR averaged in WM between C2-C5 GE Philips Siemens TR=62ms 55.80 ± 4.41 45.68 ± 2.53 48.15 ± 1.13 x
- 12. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Diffusion MRI (single subject) 12Alley et al., ISMRM 2018 A P R L 0.9 x 0.9 x 5 mm3 Fractional Anisotropy (FA) averaged in WM between C2-C5 GE Philips Siemens Issuewithsat.band&fatsat. NoPulseOx 0.65 ± 0.01 0.72 ± 0.03 0.71 ± 0.02x x
- 13. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Multi-site, multi-subjects Goals • assessing efficacy (i.e. best of the best) vs. efficiency (in situ usage) • Provide publicly-available database: could be used for generating normative values, developing new analysis methods (train models, organize challenges, etc.) Dataset • 6 subjects (3 males, 3 females), 20-40 y.o. • # sites: 6 acquired, 17 confirmed, 10 maybe —> ~200 subjects Deadline: March 31st • Results will be presented at the SC MRI workshop, Montreal, May 17th 13
- 14. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Multi-site, multi-subjects • Processing of 32 subjects done in ~1h30 (2.2GHz x 40, 512 RAM) • Example of QC: https://osf.io/z6stq/ • Preliminary results of cord CSA averaged between C2-C3 vertebral levels: 14Sites: Barcelona (Siemens-prisma_fit, n=6), Milan (Siemens-prisma, n=6), Sherbrooke (Philips-Ingenia, n=7), UCL (Philips-Ingenia, n=6), UNF (Siemens-prisma_fit, n=6) 0 15 30 45 60 75 90 2019-01-21 12:01:50 2019-01-21 12:03:28 2019-01-21 12:09:32 CSA[mm2] T2w Subjects 0 15 30 45 60 75 90 2019-01-21 11:57:39 2019-01-21 11:59:06 2019-01-21 12:05:15 CSA[mm2] Subjects T1w
- 15. Outline 1. Standardize spinal cord MRI acquisition 2. Standardize spinal cord MRI processing Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data 15
- 16. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Analysis software for spinal cord 16 problem: no standard processing tool for spinal cord brain data ?spinal cord data solution: SCT https://github.com/neuropoly/spinalcordtoolbox “SCT (Spinal Cord Toolbox) is a comprehensive and open-source library of analysis tools for multi- parametric MRI of the spinal cord.”
- 17. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Overview of SCT 17De Leener, Neuroimage 2016 Atlas-based analysis • 2D slice-by-slice • Regularized across slices & time • Robust for DWI (group-wise) Motion correction and many more features… Template and atlas C1 C3 C5 T1 Registration framework Segmentation T2w T1w SCT
- 18. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data PAM50: template of the spinal cord 18 De Leener, Neuroimage 2017 —> PAM50 Fonov, NeuroImage 2014 —> Methods for template creation Taso, MAGMA 2014 —> white matter probabilistic template Taso, NeuroImage 2015 —> white matter probabilistic template (new version) Lévy, NeuroImage 2015 —> white matter atlas T2-weighted template spinothalamic spinocerebellar corticospinal cuneatus gracilis C1 C5 T1 C3 gray matter white matter cerebrospinal fluid 0 1 0 1 Probabilistic structure White matter atlasT2-weighted template spinothalamic spinocerebellar corticospinal cuneatus gracilis C1 C5 T1 C3 gray matter white matter cerebrospinal fluid 0 1 0 1 Probabilistic structure White matter atlasSpinal cord & brainstem templates in ICBM space Benjamin De Leener Simon Lévy
- 19. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Deep learning for segmentation 19 Spinal cord segmentation • Models trained on ~3000 subjects from ~30 centers • Robust towards various pathologies Charley Gros Patient with cord compression Gros et al., arXiv:1805.06349 invivoexvivo Perone et al. Sci Rep 2018 Gray matter segmentation Christian Perone 4676 axial slices @ 100 μm isotropic
- 20. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Applications (>70 citations) Functional MRI • Kong et al. Intrinsically organized resting state networks in the human spinal cord. PNAS 2014 • Vahdat et al. Simultaneous Brain–Cervical Cord fMRI Reveals Intrinsic Spinal Cord Plasticity during Motor Sequence Learning. PLOS Biology 2015 • Eippert F. et al. Investigating resting-state functional connectivity in the cervical spinal cord at 3T. BioRxiv 2016 • Weber K.A. et al. Functional Magnetic Resonance Imaging of the Cervical Spinal Cord During Thermal Stimulation Across Consecutive Runs. Neuroimage 2016 • Eippert et al. Denoising spinal cord fMRI data: Approaches to acquisition and analysis. Neuroimage 2016 Quantitative structural MRI (diffusion, MT, etc.) • Taso et al. A reliable spatially normalized template of the human spinal cord — Applications to automated white matter/gray matter segmentation…. Neuroimage 2015 • Weber et al. Lateralization of cervical spinal cord activity during an isometric upper extremity motor task with functional magnetic resonance imaging. Neuroimage 2016 • Samson et al., ZOOM or non-ZOOM? Assessing Spinal Cord Diffusion Tensor Imaging protocols for multi-centre studies. PLOS One 2016 (in press) • Taso et al.Tract-specific and age-related variations of the spinal cord microstructure: a multi-parametric MRI study using diffusion tensor imaging (DTI) and ihMT. NMR Biomed 2016 • Massire A. et al. High-resolution multi-parametric quantitative magnetic resonance imaging of the human cervical spinal cord at 7T. Neuroimage 2016 • Duval et al. g-Ratio weighted imaging of the human spinal cord in vivo. Neuroimage 2016 Application in patients • Yiannakas et al. Fully automated segmentation of the cervical cord from T1-weighted MRI using PropSeg: Application to multiple sclerosis. NeuroImage: Clinical 2015 • Castellano et al., Quantitative MRI of the spinal cord and brain in adrenomyeloneuropathy: in vivo assessment of structural changes. Brain 2016 • Grabher et al., Voxel-based analysis of grey and white matter degeneration in cervical spondylotic myelopathy. Sci Rep 2016;6:24636. • Talbott JF, Narvid J, Chazen JL, Chin CT, Shah V. An Imaging Based Approach to Spinal Cord Infection. Seminars in Ultrasound, CT and MRI. 2016 • Martin et al. A Novel MRI Biomarker of Spinal Cord White Matter Injury: T2*-Weighted White Matter to Gray Matter Signal Intensity Ratio. AJNR 2017 • David et al. The efficiency of retrospective artifact correction methods in improving the statistical power of between-group differences in spinal cord DTI. Neuroimage 2017 • Peterson et al. Test-Retest and Interreader Reproducibility of Semiautomated Atlas-Based Analysis of Diffusion Tensor Imaging Data in Acute Cervical Spine Trauma in Adult Patients • Grabher et al. Neurodegeneration in the Spinal Ventral Horn Prior to Motor Impairment in Cervical Spondylotic Myelopathy. Journal of Neurotrauma 2017 • Smith et al. Lateral corticospinal tract damage correlates with motor output in incomplete spinal cord injury. Archives of Physical Medicine and Rehabilitation 2017 • McCoy et al. MRI Atlas-Based Measurement of Spinal Cord Injury Predicts Outcome in Acute Flaccid Myelitis. AJNR 2016 • Grabher et al., Voxel-based analysis of grey and white matter degeneration in cervical spondylotic myelopathy. Sci Rep 2016 • Hori et al., Application of Quantitative Microstructural MR Imaging with Atlas-based Analysis for the Spinal Cord in Cervical Spondylotic Myelopathy. Scientific Reports. 2018 20
- 21. Example applications of SCT
- 22. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Multi-center DTI study 22Samson, PLOS One 2016 Vanderbilt MontrealLondon A P R b=0FA 0.1 0.9 L Template Atlas London Vanderbilt M ontreal FA in fasciculus gracilis 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 L R L R L R • 3 sites • 5 subjects per site • Two MRI brands (Philips, Siemens)
- 23. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Mapping MS lesions in the cord 23Eden et al. Brain 2019 R. Bakshi (USA) C. Mainero (USA) T. Shepherd (USA) S. Smith (USA) J. Talbott (USA) D. Reich (USA) O. Ciccarelli (UK) E. Bannier (France) V. Callot (France) OFSEP (France) M. Filippi (Italy) T. Granberg (Sweden) M. Hori (Japan) K. Kamiya (Japan) Y. Tachibana (Japan) J. Talbott (USA) T. Granberg (Sweden) A. Badji (Canada) J. Maranzano (Canada) R. Zhuoquiong (China) SCTSpinal Cord Toolbox Charley GrosDominique Eden • 12 MS clinical centers —> 600 patients • 8 MD/neuroradiologists who segmented lesions • Automatic analysis with SCT
- 24. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Mapping MS lesions in the cord 24Eden et al. Brain 2019 Charley GrosDominique Eden
- 25. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data ProbabilisticdensityofMSlesions(%) Vertebral Level Mapping MS lesions in the cord 25Eden et al. Brain 2019 Vertebral Level ProbabilisticdensityofMSlesions(%) Lesion Probability Map (N=600) Charley GrosDominique Eden
- 26. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Gray matter atrophy in ALS 26Paquin et al., AJNR 17 Collaboration: PF Pradat (Pitié-Salpêtrière, Paris) Automatic GMseg. Patient #6 Patient #11 Patient #24Patient #20 Patient #23 Dice coeff. = 0.84 Dice coeff. = 0.86 Dice coeff. = 0.73 Dice coeff. = 0.70 Dice coeff. = 0.740.5 1 Manual GMseg. Image Automatic GM segmentation p-value: p=0.0041 p-value: p=0.0203 A. B. 0.0065 0.0116 0.0024 0.0707 0.0045 0.0530 0.0337 0.0412 GM atrophy is a better discriminator of ALS than cord atrophy ALSFRS-R1year PredictionError Clinical predictors Clinical predictors + SCCSA Clinical predictors + GMCSA + WM/GMCSA Mean error (Best value: 0.0) 2.05 ± 12.97 1.80 ± 8.67 1.63 ± 8.42 Prediction at 1 year Charley GrosMarie-Eve Paquin
- 27. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Shape analysis of cord compression 27Martin et al. BMJ Open 2018 Collaboration: Drs. Allan Martin & Michael Fehlings (U Toronto) • SCT provides tools to automatically analyze the shape of the spinal cord in the axial plane • Relevant metrics include antero-posterior and right-left dimensions. • Particularly interesting for studying traumatic and non-traumatic cord compression Patient with degenerative cervical myelopathy A-Pdiameter (mm) 4 5 6 7 8 9 Area(mm2) 30 40 50 60 70 Eccentricity 0.7 0.75 0.8 0.85 0.9 0.95 Symmetry 0.9 0.92 0.94 0.96 0.98 1 Orientation -10 -5 0 5 10 Superior-inferior direction (Slice #) R-Ldiameter (mm) 7 9 11 13 Metrics of spinal cord shape sensitive for cord compression
- 28. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Conclusion Spine Generic Protocol —> promote replicability, dissemination of knowledge • Minimize wasted time & $$$ spent on pilot scans for optimization • Minimize variability in multi-site, multi-vendor studies • “one-fit-all” protocol: should be adapted to specific needs • Not frozen in time: will evolve based on users’ feedback Analysis tools for spinal cord MRI —> promote reproducibility • Standardize analysis tools: more transparent, enable cross-validation of published studies. • Automatic pipelines: prevent user bias (e.g., manual delineation of ROIs), leverage large multi-center studies • Open-source analysis tools for spinal cord MRI: SCT, spinalfmri8, JIM, etc. List of software: http://www.spinalcordmri.org/software/ Communication • Discussions about unmet needs between physicists, clinicians and MRI vendors is key! 28
- 29. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Clinical Translation 29 problem: • Despite efforts in open science, these advanced acquisition/processing techniques are difficult to translate because they require expert knowledge to use them • Publishing articles is not enough • How to make them available to a larger pool of clinics worldwide? solution: Communication & Training!
- 30. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Spinal Cord MRI Workshop 30 Would you like to be on the mailing list or sponsor the event? Visit: www.spinalcordmri.org satellite workshop Toronto’15 45 attendees Spinal cord MR software Singapore’16 65 attendees Gray matter segmentation Spinal Cord MR Hack Organizers: PaulSummersUniMoRE,Modena,Italy CarloPorroUniMoRE,Modena,Italy JulienCohen-AdadPolyMTL,Montreal,Canada Registrationviaemail:paul.summers@unimore.it EventsponsoredbytheInternationalSpinalResearchTrust. Friday 16 May, 2014 12:00 – 18:00 Enterprise Hotel Corso Sempione 91 Milan, Italy Milan’14 35 attendees Standardize protocols across vendors Hawaii’17 70 attendees Paris’18 70 attendees Next workshop: Montreal: May 17th 2019 sponsored
- 31. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data www.spinalcordmri.org 31 You can find my slides there!
- 32. Cohen-Adad: Standardizing acquisition and processing of spinal cord MRI data Acknowledgements 32 Nikola Stikov Agah Karakuzu Aldo Zaimi Alexandru Foias Ariane Saliani Atef Badji Benjamin De Leener Charley Gros Christian Perone Dominique Eden Gabriel Mangeat Grégoire Germain Harris Nami Ilana Leppert Jennifer Campbell Jérôme Carretero Lucas Rouhier Matthieu Parizet Mathieu Boudreau Mélanie Lumbrano Nibardo Lopez-Rios Nicolas Pinon Ryan Topfer Stephanie Alley Tanguy Duval Tommy Boshkovski Pierre Bellec Carollyn Hurst André Cyr Gerald Moran Michael Fehlings Allan Martin Ali Akbar Claudia Gandini Wheeler-Kingshott Ferran Prados Becky Samson Francesco Grussu Armin Curt Patrick Freund Maryam Seif Shigeki Aoki Masaaki Hori Akifumi Hagiwara Koji Kamagata Kouhei Kamiya Guillaume Gilbert Suchandrima Banerjee