Dharmesh Tailor, MSE, MD, PhD1
1Radiology Specialists of Florida, Florida Hospital, Adventist Health System, Orlando, FL, United States
Synopsis
Accurate and precise mapping of metabolically
active eloquent foci in the brain, differentiating recurrent neoplasm from
pseudo-progression, differentiating tumefactive demyelination from neoplasm,
delineating the actual stroke penumbra, and early detection of
neurodegenerative disease, all require high-resolution imaging of underlying relative
brain oxygen metabolism. At the present time, however, there is no robust technique
that can map brain metabolism at a high spatiotemporal resolution. This work
derives and demonstrates a model for imaging of relative CMRO2 with
T1ρ-weighted MRI which can be readily performed on any
clinical MRI scanner with a simple pulse sequence and without the need for
hardware modification. This novel approach is optimized to yield a spatial
resolution of 0.2 mm3 and temporal resolution of 3.2 sec, and developed
to work without exogenous tracer or contrast agent administration, blood
sampling, or statistical parametrization of the image data.
Introduction
In the recent past, a few experimental MR based
techniques have been developed to image either relative or absolute regional CMRO2 in the brain
[1-4]. However, the higher resolution techniques rely on administration of an
expensive exogenous tracer, such as enriched 17O-gas, and the
techniques that do not rely of exogenous administration have been shown to
achieve an optimal 1 cm spatial resolution and an imaging time of 5.5 min for
voxel-by-voxel mapping of CMRO2 which is inadequate for most brain
mapping tasks. This study proposes a novel approach based on proton T1ρ-weighted imaging to
map relative CMRO2 in the human brain providing at least a 0.2 mm3
spatial and a 3.2 sec temporal resolution. The fast temporal resolution is
desired for avoiding motion degradation, for applications such as stroke
evaluation where time is especially critical, and for functional neuroimaging.
The sub-mm special resolution is necessary for evaluation of small mass lesions
that can best be characterized by their underlying relative metabolic activity
rather than just T2 prolongation, diffusion restriction, and contrast
enhancement. Our method does not rely on exogenous administration of a tracer
or contrast material, can be readily implemented on any clinical MR scanner
using a conventional clinical head coil without the need for any hardware
modification.
Methods
Model: Our prior data
suggests [5] that there is a linear relationship between the T1ρ-weighted
signal and PO2. If Sh represents the T1ρ-weighted
signal at a high B1 spin-locking power (at 500 Hz) and Sl
represents a low B1 spin-locking power (at 125 Hz), then the
following relationship may be derived:
$$S_h = m_h P_e+∑(i=0..n) φ(i)$$; $$S_l = m_l P_e+∑(i=0..n) φ(i) $$
In
the above equation, the sum of the function φ represents a linear combination
of other factors influencing Sh other than PO2. These would include
any contribution from T1 and T2 relaxation, as well as other MR and physiologic parameters that are independent of spin-locking
frequency. The
PO2 of a given voxel (i.e. the effective partial pressure of oxygen
or Pe) may be expressed as either a weighted linear combination of
or a weighted difference between the arterial partial pressure of oxygen (PA)
and venous partial pressure of oxygen (PV) in that voxel, such that,
$$P_e = α(P_A-P_V ) $$ and $$S_h-S_l ≡ ∆S = (m_h-m_l ) P_e$$
Letting
a new constant m = mh-ml,
$$ ∆S = mP_e = αm(P_A-P_V )$$
By
Fick’s equation, the difference between the arterial and venous oxygen content
(PA-PV) is given by the ratio of CMRO2 and
cerebral blood flow (CBF). Hence,
$$∆S = [αm/CBF] CMRO_2$$
Assuming
that CBF is a constant during the short 3.2 s total combined imaging acquisition
time of the high and low B1, the difference of high and low
frequency spin-locked T1ρ-weighted images provides a relative map of CMRO2.
Experimentation: 10 healthy human
subjects with normal conventional clinical MRIs (2 females and 8 males, ages 28
– 44) were imaged on a Siemens 3.0 T clinical scanner using a standard clinical
head coil, in accordance with an IRB approved protocol. Continuous T1ρ-weighted
images of their brains were obtained at high (500 Hz) and low (250 Hz) B1
with our sequence [6] optimized to yield acceptable SNR with ETL=60, TSL=120
ms, TR=1s, imaging time of 1.6 s, and voxel size of 0.2 mm3. No more
than 10-18% of the maximum allowable SAR was reached with this imaging at any
time, and the subjects experienced no discomfort or complications. During the
imaging acquisition, the subjects were asked to perform a motor task that also requires
the use of somatosensory/tactile perception and coordination (i.e. grabbing an
MR safe object placed by the right hip with the right hand and bringing the
object to the chin while holding the head still). The high and low spin-locked T1ρ-weighted
images obtained during performance of this task were than subtracted to yield a
relative CMRO2 image.
Results & Conclusion
A
representative subtraction T1ρ-weighted image of the high and low
spin-locking is shown in Figure 1 in the form of the raw subtraction data.
According to our model, this image provides direct visualization of relative
CMRO2. The predicted increase in metabolic activity in the left
frontal (motor) and parietal cortex (tactile perception) and the right
cerebellum (coordination), expected by the specific task design, is seen in all subjects. In conclusion, the proposed imaging approach may be
viable for high spatiotemporal resolution imaging of oxygen metabolism in the
brain.
Acknowledgements
No acknowledgement found.References
1.
Tailor
et al (2004), NeuroImage, 22, 611-18.
2.
Atkinson
et al (2010), NeuroImage, 51(2), 723-33.
3.
He
et al (2007), MRM, 57(1), 115-26.
4.
Bolar
et al (2011), MRM, 66(6), 1550-62.
5.
Tailor
et al (2009), ISMRM Proc.,17, 4527.
6.
Tailor et al, (2003),
MRM, 49(3), 479-87.