Guan Wang^{1,2}, Yi Zhang^{2}, Shashank Sathyanarayana Hegde^{2,3}, and Paul A. Bottomley^{2}

^{1}Dept. of Electrical & Computer Engineering, Johns Hopkins University, Baltimore, MD, United States, ^{2}Russell H. Morgan Dept. of Radiology & Radiological Sciences, Johns Hopkins University, Baltimore, MD, United States, ^{3}(currently) Philips Innovation Campus, Bangalore, India

### Synopsis

**Vessel
wall MRI with intravascular (IV) detectors can produce superior local
signal-to-noise ratios (SNR) and generate high-resolution T**_{1}, T_{2},
and proton density (PD) maps that could be used to automatically classify atherosclerotic
lesion stage.
However, long acquisition times potentially limit multi-parametric mapping. Here, for
the first time, spectroscopy with linear algebraic modeling (SLAM) is applied to yield accurate compartment-average T_{1}, T_{2}
and PD measures at
least 10 times faster compared to a standard full k-space reconstructed MIX-TSE sequence at 3T. Simple phase and magnitude
sensitivity corrections are incorporated
into the SLAM reconstruction to compensate for IV detector non-uniformity. ### PURPOSE

Vessel wall MRI with intravascular (IV)
detectors can produce superior local signal-to-noise ratios (SNR) [1] and high resolution
(<200μm) T

_{1}, T

_{2}, and proton density (PD) maps [2]. Combined
with machine learning-based automatic atherosclerotic lesion classification, up
to 97% area-under-curve (AUC) receiver operating characteristic (ROC) analyses
is potentially achievable [2]. However, long acquisition times are potentially
limiting for clinical T

_{1}, T

_{2} and PD mapping, and increase
the likelihood of motion artifacts. Recently, spectroscopy with linear algebraic modeling
(SLAM) was developed for MRS and CEST to reduce scan-times up to 120-fold by
delivering compartment-average signals reconstructed from a small subset of
those k-space acquisitions having the
highest SNR [3-4]. Here, for the first time, SLAM is applied to yield
compartment-average T

_{1}, T

_{2} and PD measures up to 10-fold
faster than a standard “MIX” sequence [5].
The phase and amplitude of the sensitivity profiles of the IV coils as receivers,
are incorporated into the SLAM reconstruction to compensate for intra-compartmental
non-uniformity. SLAM measurements in human vessel specimens are
compared with those from conventional full k-space
reconstruction acquired at 3T.

### METHODS

IV MRI was performed on freshly
autopsied human iliac artery specimens immersed in saline and placed in a
Philips 3T Achieva scanner with a loopless antenna in the lumen. Reference data were acquired with
an axial 2D MIX-turbo spin-echo (TSE) sequence, interleaved with inversion recovery (IR; scan-time=2.3 min; 8 echoes; voxel=0.3x0.3x5mm3; FOV=55mm; TSE-factor=16, TI=370ms, TRSE/TRIR=760/1920ms,
TE1/TE2/TE3/TE4= 25/65/105/145ms). The
T

_{1}, T

_{2} and PD values of the compartments were computed as
in [5]. Compartment-average T

_{1}, T

_{2}
and PD were measured
in compartments that were manually segmented from anatomical images (fat=F;
lesion=L; vessel fluid contents=W1; smooth vessel-wall muscle=SM, surrounding
tissue=W2) and reconstructed using: (i) the full k-space data set; and (ii) SLAM applied to the central k-space data with enough outer k-space data discarded to achieve
acceleration factors of R≤10. During
reconstruction, the magnitude and phase of the IV coil’s sensitivity were
corrected by 1/r (r=distance from the coil center; Fig.1a),
and an azimuthal phase map linearly varying from 0-2π (Fig.1b), respectively. Compartment
average T

_{1}, T

_{2} and PD values from SLAM were compared with
conventional Fourier transform (FT) MRI.

### RESULTS

Magnitude
and phase inhomogeneities of IV-MRI were successfully eliminated by the sensitivity
profile corrections (Fig.1c-d). A segmented FOV from a diseased vessel (myelodysplastic syndrome) is exemplified in Fig.2. The 10-fold accelerated SLAM
parametric maps show high consistency with the regular FT-MIX maps in Fig.3. SLAM
results for the vessel wall are plotted against R in Fig. 4, with error-bands indicating compartment mean ±standard
deviations (SD) measured in the FT-MIX maps. For R≤10, SLAM T

_{1}, T

_{2}
and PD measurements in all compartments fall within the mean±SD of the FT
results. In both lesion compartments (L1, L2) and F, and with R=10, errors in the three parameters are
≤ 0.5%(±4%) compared to the FT mean. Even in W1 and W2, where the SD of T

_{1}
and PD are ≥30% in the FT maps, SLAM T

_{1} and PD agree with the FT
means within ≤6%±6%.

### DISCUSSION

SLAM
reconstruction can be combined with IV MRI to produce accurate compartment
average T

_{1}, T

_{2}, and PD values at least R=10 times faster. In principle, R can be increased until the number of phase-encoding
steps equals the number of compartments, which in Fig. 2 would yield R=30. While the SLAM assumption of
signal uniformity within compartments [3] is adequately satisfied by applying simple
sensitivity profile corrections, these may not suffice for large R in tiny compartments with large phase variations. Nevertheless, SLAM offers a means of dramatically reducing scan-times
in high-resolution multi-parametric IV-MRI to ≤10 sec, even in regions of low SNR.
In vivo SLAM IV-MRI measurements of T

_{1}, T

_{2} and PD could enable
fast and reliable high-resolution characterization of vessel disease, and in
conjunction with automated disease classification [2], permit automatic real-time
high-resolution detection of cardiovascular disease.

### Acknowledgements

We thank Parag Karmarkar for help with the experiments and insightful discussions. This work was
supported by NIH grant R01EB007829.### References

1. El-Sharkawy AM, Qian D, Bottomley PA: The performance of interventional loopless
MRI antennae at higher magnetic field strengths. Med Phys 2008, 35:1995-2006.

2. Wang G,
Erturk MA, Hegde SS, Bottomley PA: Automated
classification of vessel disease based on high-resolution intravascular
multi-parametric mapping MRI. International
Society of Magnetic Resonance in Medicine 2015, 1659.

3. Zhang
Y, Gabr RE, Schär M, Weiss RG, Bottomley PA: Magnetic resonance Spectroscopy with Linear Algebraic Modeling (SLAM)
for higher speed and sensitivity. Journal
of Magnetic Resonance 2012, 218:66-76.

4. Zhang
Y, Gabr RE, Zhou J, Weiss RG, Bottomley PA: Highly-accelerated quantitative 2D and 3D localized spectroscopy with
linear algebraic modeling (SLAM) and sensitivity encoding. Journal of Magnetic Resonance 2013, 237:125-138.

5. Cuppen
J: RLSQ: T1, T2, and ρ calculations, combining ratios and least squares. Magnetic
resonance in medicine 1987, 5:513-524.