Simon Daniel Robinson1,2,3, Gilbert Hangel2, Beata Bachrata2, Andreas Ehrmann2, Siegfried Trattnig2, Christian Enzinger3, Markus Barth1, and Korbinian Eckstein2
1Centre for Advanced Imaging, University of Queensland, Brisbane, Australia, 2Medical University of Vienna, Vienna, Austria, 3Medical University of Graz, Graz, Austria
Synopsis
We describe a systematic optimization of gradient-echo-based
SWI at 7T, encompassing SNR in single-echo vs multi-echo acquisitions (both
monopolar and bipolar), coil combination, phase unwrapping, filtering, echo
weighting and correction of image inhomogeneity. The improvement achieved with the
resulting Contrast-weighted, Laplace-unwrapped, bipolar multi-Echo,
ASPIRE-combined, homogeneous, improved Resolution SWI (or CLEAR-SWI) is compared
to ‘Standard’ single-echo, Homodyne filtered SWI in healthy subjects and
patients. CLEAR-SWI reduces signal dropout, eliminates wrap artefacts and
provides multi-echo phase and magnitude data for R2* mapping and QSM. Applied clinically,
it provides improved visibility of multiple sclerosis lesions and the
composition of tumors.
Introduction
Susceptibility
Weighted Imaging (SWI [1,2]) uses high-pass filtered phase information
to increase the sensitivity of gradient-echo images to frequency shifts close
to susceptibility sources such as deoxygenated blood (and thereby, veins),
blood products and iron. Clinically, SWI has proved valuable in imaging tumors,
stroke, vascular dementia and multiple sclerosis. In common use, however, shortcomings
in both the acquisition and image processing lead to artefacts and suboptimal
contrast to noise ratio (CNR), particularly at ultra-high field. This study
aimed to optimize all stages of the SWI process, considering SNR and CNR achieved
with single-echo and multi-echo acquisitions (both monopolar and bipolar), coil
combination, phase unwrapping, filtering, echo weighting and correction of
image inhomogeneity. The result - Contrast-weighted, Laplace-unwrapped, bipolar
multi-Echo, ASPIRE-combined, homogeneous, improved Resolution SWI (or
CLEAR-SWI) - provides a dramatic improvement in image quality over conventional
single-echo, Homodyne-filter-generated SWI, as well as providing multi-echo phase
and magnitude data that can be used to generate R2* and Quantitative Susceptibility
Maps. All the proposed steps are computationally efficient, robust and can be
performed slice-wise, making them viable for online calculation on the scanner image
reconstruction computer. Theory
The duty cycle, or percentage of the available time that is
spent sampling the echo (rather than ramping gradients and playing out refocusing
gradients), was calculated for monopolar and bipolar acquisitions (parameters
in Fig.1 caption).
The SNR of single-echo acquisitions is given by [3]
$$\mathrm{SNR}∝\frac{2T_2^*(1 -
e^{-T_\mathrm{acq}/2T_2^*})e^{-T_\mathrm{E}/T_2^*}}{\sqrt{T_\mathrm{acq}}},$$
in which $$$T_2^*$$$ is the relaxation time, $$$T_\mathrm{E}$$$
the echo time and $$$T_\mathrm{acq}$$$ the sampling period. In multi-echo data in
which the echoes are combined using the root-sum-of-squares (RSS), the SNR can
be calculated from the SNR of the individual echoes as
$$\mathrm{SNR_{ME}}=\sqrt{\sum{\mathrm{SNR}_i^2}}.$$
Duty cycle values were used for calculation of
$$$T_\mathrm{acq}$$$ for the SNR calculation in Fig.1. Multi-echo SNR values
are relative to the achievable SNR from a single echo acquisition and therefore
exceed 100% in some cases.Methods
5 volunteers,
14 tumor patients and 1 multiple sclerosis patient were measured at 7T MAGNETOM
with a 32-channel nova medical head coil. Informed by SNR calculations (Fig.
1), a bipolar multi-echo gradient-echo sequence was used for CLEAR-SWI, with TE=4.3:4.3:25.8ms,
FA=15, BW=260Hz/px, TR=30ms, grappa=2, voxel size of 0.26x0.26x1.2mm with a
total sequence time of 9min35s. For ‘Standard’ SWI, a single echo scan was acquired
with identical parameters other than TE=19.3 and BW=60Hz/px.
For CLEAR-SWI,
phase data from the head array coil were combined using ASPIRE [4],
followed by Laplacian unwrapping [5,6], combination over echoes using
the magnitude as a weighting factor in CNR-optimal T2*-weighted combination [7].
A masked gaussian high-pass filter was applied in 2D with a sigma of 4 pixels
and the phase mask generated using the sigmoidal function $$$f(x)=(1+\mathrm{tanh}(1-x/m))/2$$$
(Fig.2c). Magnitude data were combined over echoes using i) root-sum-of-squares
(RSS) combination to optimize SNR and reduce signal dropouts and ii) CNR-weighted
combination to optimize the CNR between two chosen tissue types. After
homogeneity correction of the combined magnitude [8], the phase mask and
magnitude were multiplied to generate the SWI.
For Standard
SWI, the manufacturer’s online Homodyne filtering-based reconstruction was used. Results
Our model of duty
cycle/sampling efficiency showed advantages for bipolar acquisitions,
particularly for short echo spacings and high resolution. For monopolar
acquisitions and the example parameters in Fig. 1 (4 ms echo spacing, 220 mm
FoV), less than 50% of the TR was spent acquiring signal above a matrix size of
448 (0.5mm resolution), with the remainder of the time being spent on fly-back gradients.
For bipolar acquisitions the efficiency was, for all resolutions, above 90% and
the SNR was about 30% higher than in single echo acquisitions. The sigmoid phase
filter we propose reduces background noise and increases vessel contrast
(Fig.2). The homogeneity filter we adopted (Fig.3) was effective in removing
coil sensitivity effects. Fig. 4 illustrates the main features of CLEAR-SWI
compared to Standard SWI: signal dropouts are reduced by the weighting of early
echoes in those regions, phase artefacts are eliminated by the use of Laplacian
unwrapping and the SWI is more homogeneous, making it easier to window and read.
In imaging tumors, CLEAR-SWI showed no
phase wraps or signal dropouts close to pathological tissue and the additional
M0-map (the initial signal in the T2* fit, which is available due to using a
multi-echo sequence) helped to identify the tumor boundaries (will be
presented). In the patient with multiple sclerosis, all lesions were either more
clearly visible on CLEAR-SWI than Standard SWI (e.g. Lesion 1 in Fig.5) or only
visible on CLEAR-SWI (e.g. Lesion 2 in Fig.5).Conclusions
CLEAR-SWI combines
optimized methods for each step of the SWI process to produce images which are
dramatically improved compared to conventional, single-echo, Homodyne filtered
SWI. CLEAR-SWI reduces signal dropout, increases SNR and CNR and eliminates
wrap artefacts and inhomogeneities, generating images which provides additional
insights into the composition of tumors and clinically relevant features in MS.
In future work, CLEAR-SWI will be implemented on the image reconstructor to provide
SWI for clinicians and T2* and QSMs for researchers.Acknowledgements
This study was
funded by the Austrian Science Fund project FWF31452. SR was supported by the Marie
Skłodowska-Curie Action MS-fMRI-QSM 794298.References
[1] Haacke, E.M., Xu, Y., Cheng, Y.-C.N.,
Reichenbach, J.R., 2004. Susceptibility weighted imaging (SWI). Magnetic resonance in
medicine 52, 612–618.
[2] Reichenbach,
J.R., Venkatesan, R., Schillinger, D.J., Kido, D.K., Haacke, E.M., 1997. Small
vessels in the human brain: MR venography with deoxyhemoglobin as an intrinsic
contrast agent. Radiology 204, 272–277.
[3] Jutras, J.-D., Wachowicz, K., Gilbert, G., De
Zanche, N., 2017. SNR
efficiency of combined bipolar gradient echoes: Comparison of three-dimensional
FLASH, MPRAGE, and multiparameter mapping with VFA-FLASH and MP2RAGE. Magnetic
Resonance in Medicine 77, 2186–2202.
[4]
Eckstein, K., Dymerska, B., Bachrata, B., Bogner, W., Poljanc, K., Trattnig,
S., Robinson, S.D., 2018. Computationally Efficient Combination of
Multi-channel Phase Data From Multi-echo Acquisitions (ASPIRE). Magnetic
Resonance in Medicine 79, 2996–3006.
[5]
Schofield, M.A., Zhu, Y., 2003. Fast phase unwrapping algorithm for
interferometric applications. Opt. Lett., OL 28, 1194–1196.
[6] Rauscher,
A., Sedlacik, J., Deistung, A., Mentzel, H.J., Reichenbach, J.R, 2006. Susceptibility
Weighted Imaging: data acquisition, image reconstruction and clinical applications.
Z Med Phys. 16(4):240-50.
[7] Wu, B., Li, W., Avram, A.V., Gho, S.-M., Liu, C.,
2012. Fast and
tissue-optimized mapping of magnetic susceptibility and T2* with multi-echo and
multi-shot spirals. NeuroImage, Neuroergonomics: The human brain in action and
at work 59, 297–305.
[8] Eckstein, K., Trattnig, S., Robinson, S.D., 2019. A Simple Homogeneity
Correction for Neuroimaging at 7T, in: Proceedings of the 27th Annual Meeting
ISMRM. Presented at the ISMRM, Montréal, Québec, Canada, 2011.