Clinical · Neurology MAGNETOM Flash (85) 3/2023
Advancing Clinical and Neuroscientific
Research Through Accessible and
Optimized Protocol Design at 3T
Sriranga Kashyap, Ph.D.1; Kâmil Uludağ, Ph.D.1,2,3,4
1
BRAIN-TO Lab, Krembil Brain Institute, University Health Network, Toronto, ON, Canada
2
Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada
3
Center for Neuroscience Imaging Research, Institute for Basic Science, Suwon, Republic of Korea
4
Department of Biomedical Engineering, Sungkyunkwan University, Suwon, Republic of Korea
Background projects. Even sequences considered fairly routine and
standard, such as T1w MPRAGE, have a large variation in
Rising concern about healthcare costs and research
imaging parameters, and there is no guarantee they will
funding is driving transformative shifts in the field of MR
be compatible with post-hoc data analyses. In this article,
imaging. Collaborative efforts among academics, clini-
we detail a strategic approach to developing a set of
cians, and industry experts are helping dispel outdated
optimized neuroimaging protocols and present optimized
notions of MRI as a slow and costly modality. Nevertheless,
parameters for the 3T MAGNETOM Prisma scanner
cutting-edge technological advancements achieved in
(Siemens Healthcare, Erlangen, Germany) that are not
academia and industry are difficult to translate into a shift
only faster and more efficient but have also been tested
in cognitive neuroscience and clinical research and patient
and validated using community-standard MRI post-
care due to a multitude of factors – particularly the avail-
processing software.
ability of technical expertise and time. Clinicians and
neuroscientists often rely on colleagues and/or consortia
(such as the Human Connectome Project and the Design principles
Alzheimer’s Disease Neuroimaging Initiative) to obtain
When optimizing the BRAIN-TO (Brain Research in
imaging protocols for their research. This results in a wide
Advanced Imaging and Neuromodeling – Toronto) proto-
variation in imaging protocols used within a single institu-
cols, several overarching principles were applied:
tion, and even by a single researcher for their different
Structural Imaging Diffusion Imaging
Perfusion Imaging BOLD Imaging
1 Example BTO imaging strategies laid out using the MAGNETOM Prisma Dot Cockpit interface.
2 siemens-healthineers.com/magnetom-world
MAGNETOM Flash (85) 3/2023 Neurology · Clinical
First, the protocols should be built on product sequences BRAIN-TO (BTO) protocols
from Siemens Healthineers (no work-in-progress packages
All data acquired from BTO protocols have been tested and
or C2Ps) to maximize accessibility.
validated for clinical and neuroscientific research applica-
Second, all protocols should have spatially isotropic
tions using community-standard tools such as Freesurfer
voxels to reduce partial voluming and voxel volumes.
[1], FSL [2], SPM [3], CAT [4], and AFNI [5]. Generalized
Because the cortex is curved, using anisotropic voxels
versions of the code and/or containers, as well as a wiki
(as is often the case in patient care) potentially reduces
with recommendations for post-processing, are available
diagnostic value in the dimension with the lowest spatial
through our lab at https://github.com/BRAIN-TO/bto_mri_
resolution and should therefore be avoided.
protocols_info.
Third, each scan should have a maximum acquisition
time of 5 to 6 minutes (exceptions include multi-echo
FLASH, multi-shell diffusion, resting-state, and functional Structural imaging
MRI) to minimize participant discomfort and motion
In line with the design principles outlined earlier, the struc-
artifacts.
tural imaging component of the BRAIN-TO protocol set
Fourth, the protocols should be optimized to take
constitutes 1 mm isotropic 3D MPRAGE, 3D MP2RAGE, 3D
advantage of the acceleration capabilities of the head coil
SPACE T2, and 3D SPACE FLAIR sequences. A multi-echo
(20-channel Head/Neck, 32-channel Head, 64-channel
FLASH protocol has also been incorporated. It enables
Head/Neck), thus offering higher spatial and/or temporal
quantitative T2* mapping, susceptibility weighted imaging
resolution.
(SWI) [6], and quantitative susceptibility mapping (QSM)
Finally, these optimized protocols should be organized
[7] and takes about 10 minutes to acquire in 1 mm iso
using the Strategy and Decision Tree features in the Dot
tropic and high-resolution variants (only tested for 32-
Cockpit interface from Siemens Healthineers. This means
and 64-channel coils). It is important to point out that
they will offer easy access to the end-users and can be
clinical research applications have different requirements
made available to collaborating institutes as .exar1 pack
to diagnostic imaging [8]. In this context, the increase
ages. Unless otherwise specified, the parameters, data,
in resolution and savings in time are a good trade-off [9]
and results illustrated hereafter are for the 20-channel
to enable resolution and field-of-view (FOV) matching
Head/Neck coil.
of the T1w and FLAIR images and their integration into
a post-processing pipeline without resampling artifacts.
Spatial Acquisition Acquisition Voxel
Sequence resolution Sequence parameters time time volume
(mm) (min:sec) difference difference
GRAPPA = 2, TE = 4.1 ms, TR = 2000 ms, IR = non-sel,
Typical 3D
1×1×1 TI = 899 ms, α = 8°, FOV = 256 × 256, 160 axial slices, 06:00
T1w MPRAGE
Inline MPR = off
GRAPPA = 2, TE = 2.88 ms, TR = 2100 ms, IR = non-sel,
BTO 3D
1×1×1 TI = 900 ms, α = 9°, FOV = 256 × 256, 256 sagittal slices, 04:20 – 28% 0%
T1w MPRAGE
Inline MPR = tra
GRAPPA = 2, TE = 94 ms, TR = 9000 ms, TI = 2500 ms,
Typical IR = Slice-sel, α = 150°, FOV = 275 × 245, 32 axial slices,
IR 2D TSE 0.8 × 0.8 × 4 02:44
FLAIR distance factor = 25%, Inline MPR = off,
interpolation = on
CAIPIRINHA = 2 × 2, TE = 393 ms, TR = 4500 ms,
BTO
3D SPACE 1×1×1 TI = 1800 ms, IR = non-sel, α = variable, 04:14 + 55% – 61%
FLAIR
FOV = 256 × 256, 256 sagittal slices, Inline MPR = tra
GRAPPA = 2, TE = 2.98 ms, TR = 4500 ms,
BTO
3D TI1/TI2 = 850/2550 ms, IR = non-sel, α1/α2 = 5°/6°,
T1 1×1×1 07:11 + 20% 0%
MP2RAGE FOV = 256 × 256, 224 sagittal slices,
map
distance Inline MPR = tra, MapIt = T1 map
BTO GRAPPA = 2, TE1-3 = 12/24/38 ms, TR = 55 ms, α = 20°,
3D GRE 1×1×1 10:35
multi-echo FOV = 224 × 224, 160 axial slices
Table 1: Comparison of typical vs. BTO protocols for structural imaging
siemens-healthineers.com/magnetom-world 3
Clinical · Neurology MAGNETOM Flash (85) 3/2023
T1w FLAIR SWI min IP
T2w Freesurfer QSM T2*
2 Example images from the BTO structural protocols and their post-processed results.
Diffusion imaging
The diffusion imaging component consists of a fast single- as FSL and MRtrix3 [12]. Two example scenarios can be en-
shell protocol, a comprehensive single-shell protocol, and visioned here. In the first, the investigator is only interested
a multi-shell protocol for neurite orientation dispersion in assessing fractional anisotropy (FA), mean diffusivity
density imaging (NODDI) [10]. Each protocol has been (MD), and diffusion tensor (DTI). In that case, they can
implemented with two versions: first, with the default opt for the BTO single-shell, fast protocol to acquire the
diffusion vector set from Siemens Healthineers; second, data in 3 minutes. In the second scenario, the researcher
with a custom diffusion vector set with interspersed b0s is also interested in tractography, and can opt for the BTO
while ensuring uniform coverage of q-space [11]. The data single-shell protocol to acquire the data in 5.5 minutes.
acquired from BTO protocols were tested, processed, and The end-user can make these decisions with the knowl-
evaluated for quality using community-standard tools such edge that the data will be of sufficient quality during
Spatial Acquisition Acquisition Voxel
Sequence resolution Sequence parameters time time volume
(mm) (min:sec) difference difference
GRAPPA = 2, SMS = 2, TE = 68 ms, TR = 3400 ms,
Partial Fourier = 6/8, FOV = 230 × 230,
Typical
2D EPI 2.4 × 2.4 × 2.4 echo-spacing = 0.76 ms, bandwidth = 1488 Hz/px, 07:37
DWI
68 axial slices, diff. directions = 64,
b1/b2 = 0/1000 s/mm2, averages b1/b2 = 1/2
GRAPPA = 2, SMS = 2, TE = 75 ms, TR = 4500 ms,
BTO
Partial Fourier = 6/8, FOV = 220 × 220,
DWI
2D EPI 2×2×2 echo-spacing = 0.56 ms, bandwidth = 2164 Hz/px, 05:33 – 27% – 42%
(single-
84 axial slices, diff. directions = 64,
shell)
b1/b2 = 0/1000 s/mm2, averages b1/b2 = 5/1
GRAPPA = 2, SMS = 2, TE = 75 ms, TR = 4500 ms,
BTO
Partial Fourier = 6/8, FOV = 220 × 220,
DWI
2D EPI 2×2×2 echo-spacing = 0.56 ms, bandwidth = 2164 Hz/px, 03:00 – 61% – 42%
(single-
84 axial slices, diff. directions = 30,
shell, fast)
b1/b2 = 0/1000 s/mm2, averages b1/b2 = 5/1
08:33
GRAPPA = 2, SMS = 2, TE = 75 ms, TR = 4500 ms,
As 2 runs
BTO Partial Fourier = 6/8, FOV = 220 × 220,
Run 1: 30 dirs,
NODDI 2D EPI 2×2×2 echo-spacing = 0.56 ms, bandwidth = 2164 Hz/px, + 12% – 42%
b1000 = 03:00
(multi-shell) 84 axial slices, diff. directions = 30/64,
Run 2: 64 dirs,
b1/b2 = 0/1000/2000 s/mm2, averages b1/b2 = 5/1/1
b2000 = 5:33
Table 2: C
omparison of typical vs. BTO protocols for diffusion imaging
4 siemens-healthineers.com/magnetom-world
MAGNETOM Flash (85) 3/2023 Neurology · Clinical
CNRb1000 CNRb2000 FA MD V1
Viso Vec Vic NDI ODI
3 Example images from the BTO diffusion NODDI protocol and its post-processed results.
post-processing. It is important to reiterate that these
are optimized, general-purpose protocols. They are not
designed for bespoke investigations.
Perfusion imaging
Non-invasive perfusion MRI is typically carried out using
Arterial Spin Labelling (ASL) sequences that use mag
netically labelled arterial blood water as an endogenous
tracer to measure blood flow. While community-standard
protocols [13] with spatial resolutions of 3 to 4 mm
in-plane and 4 to 8 mm slice thickness may suffice for
detecting macroscale perfusion patterns in the brain,
the anisotropicity and increased partial voluming is sub-
optimal for clinical and neuroscientific research. The BTO
perfusion protocol set consists of whole-brain 3 mm and
2.5 mm isotropic pseudocontinuous ASL (pCASL) and
pulsed ASL (PASL) options in a total acquisition time of
4 to 6 minutes. Recently, we pushed this boundary further
to evaluate the feasibility of 2 mm isotropic whole-brain
ASL and the impact of coil and post-processing choices on
the perfusion maps [14].
0 mL / 100 g / min 100
BOLD imaging (functional MRI)
The functional MRI (fMRI) component that uses the xample slices of a single-subject cerebral blood flow map from
4 E
gradient-echo 2D EPI sequence from Siemens Healthineers the BTO 2.5 mm isotropic perfusion protocol.
siemens-healthineers.com/magnetom-world 5
Clinical · Neurology MAGNETOM Flash (85) 3/2023
is by far the most diverse component of the BTO protocol BTO BOLD protocol has an increased SMS factor of 4
set. Particular emphasis has been placed on providing and a volume TR of 1 second for the same spatial
investigators with a modern fMRI protocol with 2.4 mm resolution. While staying true to the design principles and
isotropic resolution and a community-standard TR of generalizability, the BTO BOLD protocols also offer direct
2.0 seconds employing nominal acceleration factors and decision-making choices without having to worry about
applicable for any coil. All BTO fMRI protocols have dedi the implications of the other parameter choices, as
cated variants for 32- and 64-channel coils to make maxi- we have assessed the protocols using both resting and
mum use of the acceleration capabilities. This results in task fMRI. We are delighted to report that, over the past
particularly high in-plane resolution and/or Simultaneous year, our protocol offerings have covered a vast array
Multi-Slice (SMS) factors for higher temporal resolution. of requirements at our site and did not require case-by-
For example, the 64-channel coil variant of the standard case modifications.
Spatial Acquisition Acquisition Voxel
Sequence resolution Sequence parameters time time volume
(mm) (min:sec) difference difference
GRAPPA = 2, TE = 30 ms, α = 85°, FOV = 220 × 220,
Typical
2D EPI 3×3×4 echo-spacing = 0.93 ms, bandwidth = 1184 Hz/px, 2.0
BOLD
40 axial slices
GRAPPA = 2, SMS = 2, TE = 30 ms, α = 70°,
BTO BOLD
2D EPI 2.4 × 2.4 × 2.4 FOV = 220 × 220, echo-spacing = 0.49 ms, 2.0 – 62%
(std)
bandwidth = 2470 Hz/px, 68 axial slices
BTO BOLD GRAPPA = 2, SMS = 2, TE = 30 ms, α = 70°,
(std 2D EPI 2×2×2 FOV = 220 × 220, echo-spacing = 0.53 ms, 2.0 – 78%
high-res) bandwidth = 2272 Hz/px, 62 axial slices
32- / 64-channel coils
RAPPA = 2, SMS = 4, TE = 30 ms, α = 70°,
BTO BOLD
2D EPI 1.6 × 1.6 × 1.6 FOV = 220 × 220, echo-spacing = 0.61 ms, 2.0 – 88%
(high-res)
bandwidth = 1906 Hz/px, 100 axial slices
GRAPPA = 2, SMS = 4, TE = 30 ms, α = 70°,
BTO BOLD
2D EPI 2.4 × 2.4 × 2.4 FOV = 220 × 220, echo-spacing = 0.49 ms, 1.0 – 50% – 62%
(fast)
bandwidth = 2470 Hz/px, 68 axial slices
Table 3: Comparison of typical vs. BTO protocols for BOLD imaging
Mean EPI tSNR xample slices of the
5 E
mean EPI images and
temporal signal-to-noise
2 mm iso ratio (tSNR) of three
TR = 2 s BTO BOLD fMRI protocols
Default Mode Network acquired using the
(1.6 mm iso rs-fMRI) 32-channel Head coil.
Illustration of the Default
Mode Network obtained
(hires) from a high-resolution
1.6 mm iso (hi-res) 1.6 mm isotropic
TR = 2 s resting-state BOLD fMRI
acquisition.
(fast)
2.4 mm iso
TR = 1 s
6 siemens-healthineers.com/magnetom-world
MAGNETOM Flash (85) 3/2023 Neurology · Clinical
Summary 4 Eickhoff SB, Stephan KE, Mohlberg H, Grefkes C, Fink GR, Amunts K,
et al. A new SPM toolbox for combining probabilistic cytoarchitec-
The BRAIN-TO (BTO) protocols present a comprehensive ap- tonic maps and functional imaging data.
proach to making optimized neuroimaging protocols acces- NeuroImage. 2005;25(4):1325–35.
sible to clinical and neuroscientific researchers. The BTO 5 Cox RW. AFNI: Software for Analysis and Visualization of Functional
Magnetic Resonance Neuroimages.
protocols are designed based on key principles, utilizing
Comput Biomed Res. 1996;29(3):162–73.
Siemens Healthineers product sequences for accessibility. 6 Eckstein K, Bachrata B, Hangel G, Widhalm G, Enzinger C, Barth M,
They incorporate spatially isotropic voxels, efficient scan et al. Improved susceptibility weighted imaging at ultra-high field
times, and dedicated variants for using head coils with using bipolar multi-echo acquisition and optimized image
denser coil arrays. We hold Town Hall sessions to dissemi- processing: CLEAR-SWI. NeuroImage. 2021;237:118175.
7 Langkammer C, Bredies K, Poser BA, Barth M, Reishofer G, Fan AP,
nate information about the BTO protocols to our on-site
et al. Fast quantitative susceptibility mapping using 3D EPI and total
neuroimaging community. To improve the accessibility to generalized variation. NeuroImage. 2015;111:622–30.
advanced neuroimaging, we make the guidelines for pro- 8 Kakeda S, Korogi Y, Hiai Y, Ohnari N, Sato T, Hirai T. Pitfalls of 3D
cessing data acquired with the standard protocols and FLAIR brain imaging: a prospective comparison with 2D FLAIR.
analysis code available through a dedicated GitHub reposi- Acad Radiol. 2012;19(10):1225–32.
9 Kitajima M, Hirai T, Shigematsu Y, Uetani H, Iwashita K, Morita K,
tory (https://github.com/BRAIN-TO/bto_mri_protocols_info).
et al. Comparison of 3D FLAIR, 2D FLAIR, and 2D T2-Weighted MR
The BTO protocols not only cover diverse imaging modali- Imaging of Brain Stem Anatomy.
ties, including structural imaging, diffusion imaging, perfu- AJNR Am J Neuroradiol. 2012;33(5):922–7.
sion, and BOLD imaging, but have also been rigorously 10 Zhang H, Schneider T, Wheeler-Kingshott CA, Alexander DC. NODDI:
tested and validated using community-standard post-pro- Practical in vivo neurite orientation dispersion and density imaging
of the human brain. NeuroImage. 2012;61(4):1000–16.
cessing tools. This enhances their reliability, reassures the
11 Caruyer E, Lenglet C, Sapiro G, Deriche R. Design of multishell
end-user, and makes cutting-edge plug-and-play neuroim- sampling schemes with uniform coverage in diffusion MRI:
aging a reality. Design of Multishell Sampling Schemes.
Magn Reson Med. 2013;69(6):1534–40.
12 Tournier JD, Smith R, Raffelt D, Tabbara R, Dhollander T, Pietsch M,
et al. MRtrix3: A fast, flexible and open software framework for
medical image processing and visualisation.
NeuroImage. 2019;202:116137.
13 Alsop DC, Detre JA, Golay X, Gunther M, Hendrikse J,
References Hernandez-Garcia L, et al. Recommended implementation of
arterial spin-labeled perfusion MRI for clinical applications:
1 Fischl B. FreeSurfer. NeuroImage. 2012;62(2):774–81.
A consensus of the ISMRM perfusion study group and the European
2 Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TEJ,
consortium for ASL in dementia.
Johansen-Berg H, et al. Advances in functional and structural MR
Magn Reson Med. 2015;73(1):102–16.
image analysis and implementation as FSL.
14 Kashyap S, Oliveira ÍAF, Uludağ K. Feasibility of high-resolution
NeuroImage. 2004;23 Suppl 1:S208–19.
perfusion imaging using Arterial Spin Labelling MRI at 3 Tesla.
3 Ashburner J. Computational anatomy with the SPM software.
biorXiv [Internet]. 2023 [cited 2023 Sep 2]; Available from:
Magn Reson Imaging. 2009;27(8):1163–74.
http://biorxiv.org/lookup/doi/10.1101/2023.08.02.551576
Contact
Sriranga Kashyap, Ph.D.
Research Fellow, BRAIN-TO Lab
Krembil Brain Institute, Toronto Western Hospital
University Health Network
135 Nassau St
Toronto, ON, M5T 1M8
Canada
Kâmil Uludağ, Ph.D. Sriranga Kashyap, Ph.D. sriranga.kashyap@uhn.ca
siemens-healthineers.com/magnetom-world 7
