Hong Shang1,2, Subramaniam Sukumar1, Robert A. Bok1, Irene Marco-Rius1, Cornelius von Morze1, Adam B. Kerr3, Galen Reed4, Michael Ohliger1, John Kurhanewicz1, Peder E. Z. Larson1, John M. Pauly3, and Daniel B. Vigneron1
1Radiology and Biomedical Imaging, UCSF, San Francisco, CA, United States, 2Bioengineering, UC Berkeley - UCSF, San Francisco / Berkeley, CA, United States, 3Electrical Engineering, Stanford University, Stanford, CA, United States, 4HeartVista, Menlo Park, CA, United States
Synopsis
Balanced SSFP
sequences can provide superior SNR efficiency for hyperpolarized 13C
imaging, by efficiently utilizing the non-recoverable magnetization. A
spectrally selective bSSFP sequence was developed to enable fast mapping of
hyperpolarized metabolites. A novel approach for bSSFP spectral selectivity was
developed utilizing a combination of optimized multiband RF pulses and a bSSFP pulse
train with a carefully chosen TR to avoid banding artifact. The sequence enabled
3D dynamic imaging of HP resonances generated in studies with co-polarized pyruvate
and urea (with ~1% selectivity), attaining 2mm isotropic resolution and <1s
temporal resolution.Purpose
Balanced SSFP
(bSSFP) sequences can provide superior SNR efficiency for hyperpolarized (HP)
13C
imaging, by efficiently utilizing the non-recoverable magnetization. It has
been applied for HP
13C perfusion imaging and angiography using
metabolically inactive agents with a single resonance
1,2. In this work, a
spectrally selective bSSFP sequence was developed to enable metabolic imaging
of multiple HP resonances. The sequence was optimized for dynamic 3D imaging of
individual resonances in experiments with
co-polarized [1-
13C]pyruvate
and [
13C]urea, attaining 2mm
isotropic spatial resolution and <1s temporal resolution.
Methods
Unique
challenges faced in HP 13C metabolite mapping include the non-recoverable
nature of magnetization and the complex pattern of resonances that must be
spectrally resolved. A novel approach for bSSFP spectral selectivity was
developed utilizing a combination of optimized multiband RF pulses and a bSSFP
pulse train with a carefully chosen TR to avoid banding artifact. This provides
the flexibility to separately excite injected substrates and metabolic products
with different flip angles (FA), which is key for dynamic studies 3.
bSSFP has an
intrinsic periodic frequency response, as shown in Fig.1 (A, B). Spins near the
center of each cycle behave as desired, in that a large FA excites
magnetization while a small FA has little affect. The opposite behavior exists
at the edge of each cycle; where a large FA leaves little longitudinal and
transverse magnetization while a small FA produces significant excitation over
a small frequency range. When a selective RF pulse is used (Fig.1 (C)), besides
the predefined central passband, there are also narrow excitation bands at the edge
of each cycle due to accumulated small FA, as shown in Fig.1(D). A new method
is proposed to avoid this banding artifact by carefully choosing TR such that the
resonant frequency of each compound is located near center of bSSFP frequency
response cycles, thus driving the overall spectral selectivity close to the
spectral selectivity of RF pulse.
An optimized multiband
spectrally selective RF pulse with shortest possible duration 4 was designed to
keep TR and total scan time short, which is crucial due to the rapid
respiratory motion of mice imaged in this study, which necessitates sub-second
imaging with respiratory triggering. The minimum pulse duration was achieved by
exploiting spectral sparsity and releasing the constraint on “don’t-care”
regions. The optimal pyruvate/lactate/urea-only pulse has duration of
1.4ms/2.16ms/1.04ms, much shorter than conventional minimum-phase SLR pulses
(~6ms), with one example shown in Fig. 3 (C, D).
A ramp-up
preparation pulse was applied before acquisition to reduce transient state
signal oscillations. A ramp-down pulse was applied after the acquisition to
flip magnetization back to Mz to increase signal, as shown in Fig. 3(A).
Results
Bloch simulation
of bSSFP signal evolution around one stopband (urea) is shown in Fig.2, which
failed with a sub-optimal TR of 4ms (B), but succeeded with an optimal TR of
3.8ms (D).
Pyruvate
acquisitions with single compound phantom at different resonant frequencies are
shown in Fig. 4 (A). Ideally, signal should only be seen in the pyruvate image,
as shown in the bottom row with TR of 3.8ms. Such selectivity cannot be
achieved when TR is slightly off, such as the top row with TR of 4ms. Note that the failed selectivity at
urea/alanine/lactate band agrees with the simulation results in Fig.2 (A) where
those bands touch the banding artifact region. Additionally, pyruvate/lactate/urea
acquisitions of HP pyruvate/lactate/urea solution are shown in Fig.4 (B), with a
table of measured spectral selectivity.
In vivo 3D dynamic bSSFP acquisitions of pyruvate/lactate/urea
on a normal mouse with injected co-polarized pyruvate/urea are shown in Fig. 5.
For injected substrate pyruvate/urea, strong signal was observed in the aorta at
early time points. Lactate, converted from pyruvate, was also observed, mostly in
kidney.
Discussion and Conclusion
A spectrally
selective 3D dynamic bSSFP sequence was developed for HP 13C metabolite
mapping. The spectral selectivity (~1%) of the bSSFP sequence was achieved by
using optimal RF pulses with shortest duration and carefully choosing TR to
avoid banding artifacts.
Compared to
other HP 13C imaging sequences, such as spoiled gradient-echo
sequences with EPI or EPSI readouts 3, bSSFP is more SNR efficient. Compared
to spin echo sequences 5, bSSFP also has the refocusing performance but can be
used in the small FA regime to prolong the observation time window. Compared to
multi-echo spectrally selectively bSSFP sequences 6, this sequence features
easy and robust reconstruction, and the flexibility of exciting different compounds
with different FA. Compared to low FA spectrally selectively bSSFP sequences 7,8,
this sequence yields higher SNR with larger FA and also wider passbands.
Acknowledgements
This work was supported by NIH grants P41EB013598.References
[1] Svensson et
al. MRM 50.2: 256-262 (2003), [2] Reed et al. IEEE TMI 33: 362-371 (2014), [3] Larson
et al. JMR 194.1:121-127 (2008), [4] Shang et al. ISMRM (2015), [5] Cunningham
et al. JMR 187.2: 357-362 (2007), [6] Leupold
et al. MAGN RESON MATER PHY 22.4: 251-256 (2009). [7] von Morze et al. MRI
31.2: 163-170 (2013). [8] Månsson et al. MRM 68.6: 1894-1899 (2012).