Charles Cantrell1, Yong Jeong2, Kevin Midlash3, Parmede Vakil2, Sameer Ansari2, and Timothy J Carroll3
1Northwestern University, Chicago, IL, United States, 2Northwestern University, 3University of Chicago
Synopsis
The
length scales associated with parenchymal oxygen extraction fraction (OEF),
coupled with the near uniformity of normative OEF across the brain dictate the
development of an imaging approach that is sensitive to low-spatial frequency
imaging behavior. Previous approaches to measure OEF using MRI have utilized
high pass-filters—effectively removing much of the signal. We propose a new method to filter out
geometric field inhomogeneity, by imaging temporally through the cardiac
cycle. In a study in 11 patients with intracranial
atherosclerotic disease, we found elevated OEF on the compromised hemisphere as
compared to the healthy contralateral side (p<0.0195).
Introduction
Over
the past 20 years, considerable work has been done to understand the brain’s
auto-regulatory systems and how certain deficiencies may lead to hemodynamic
failure. By quantifying compensatory
mechanisms, Derdeyn found that patients with increased cerebral oxygen
extraction fraction (OEF) distal to a carotid artery occlusion, were more
likely to have a stroke within the next year [1]. Traditional approaches
developed to measure cerebral OEF with MRI rely on the susceptibility
differences between oxygenated and deoxygenated blood. The length scales
associated with parenchymal OEF, coupled with the near uniformity of normative
OEF across the brain dictate the development of an imaging approach that is
sensitive to low-spatial frequency imaging behavior. Removal or “normalization”
of low spatial frequency signal is required to mitigate air/tissue boundaries
in the auditory canals and frontal sinus and other sources of non-uniform local
magnetic fields. The residual phase resulting from geometric field
inhomogeneity has been a challenge for the clinical translation of OEF
imaging. To address this challenge, we
have developed scan protocol and post processing algorithm to filter out
geometric field inhomogeneity and relax the spatial frequency requirements on the
susceptibility mapping used in OEF imaging.
Outside
the static dephasing regime, magnetic shift between oxygenated hemoglobin (oHb)
and deoxygenated hemoglobin (dHb) creates a local frequency and phase shift
proportional to OEF[2]: $$δω = 4/3 γ π Δx_0 * Hct * (OEF * B_0)$$
However local alterations in the main magnetic field $$$(ΔB_0 (x,y) $$$ are large
(~100-200 Hz at 3.0T) and overwhelm the 15–30 Hz signal changes anticipated
from theory[2]. To mitigate the effect of static field inhomogeneity, we
synchronize our acquisition to the inflow of oxygenated blood and integrate the
change in susceptibility over the cardiac cycle. Methods
Cerebral Oxygen Extraction Fraction: Twieg’s Parameter Assessment by Retrieval
from Signal Encoding (PARSE) acquisition[3] is based on a rosette trajectory and maximum likelihood fit yields parametric
images of frequency offset, R2* and M0. PARSE provides high,
accurate sensitivity to local frequency shits in an 80 ms readout and is
therefore ideal for OEF imaging(Figure 1). We acquire 20 single slices acquired
at successive times after the cardiac QRS complex. The goal is to measure the
change in frequency shift as oxygenated blood is metabolize into de-oxygenated
blood. The dynamic signal changes
acquired with PARSE were decomposed into d.c. components (Figure 2) which
contain all the local magnetic field inhomogeneity and dynamic components which
represent the change in local susceptibility which occur as oxygen is extracted
from fresh blood supplied to the parenchyma:
$$ δω(x,y,t) = γΔB_0(x,y) + ∫4/3 γ π Δx_0 * Hct * (OEF(x,y,t) * B_0) dt $$ Cerebral
Reactivity: Dynamic
components of the OEF signal (OEF(x,y,t)) were modelled a Linear Time Invariant System where the input
function (i.e. heartbeat) is approximated as a unit impulse and the change in
parenchymal susceptibility after the initial delivery of oxygenated blood is
the system output. We define the Vascular Reactivity Function(VRF): VRF ≈ 1/Vr e(-t/V_rC) where Vr is the vascular
resistance of the brain and C is the vascular compliance, characterizes the
system. We define b
= 1/VrC to be
vascular reactivity.
The PARSE acquisition consisted of a single slice, 5.0mm thick, 210mm x 210mm FOV, 108x108 matrix, resolution = 1.94x1.94x5
mm3 2D PARSE images.
Prospective cardiac gated was implemented in patients to synchronize
image acquisition to the inflow of oxygenated blood. Each 2D slice was acquired
25 times, at 25 ms increments from the R-trigger throughout the cardiac cycle
(25 ms to 625 ms delay).
Results
A series of
11 consecutive patients (M/F 5/6, <age>=52.1±11.1) referred for the
evaluation of symptoms indicative of ischemic
stroke or transient ischemic attack (TIA), secondary to intracranial
atherosclerotic disease (ICAD). OEF in
the normal hemispheres was (44%±6.7%) which is in agreement with historical
reference PET-OEF and significantly increase (56 %±6.7%, p<0.0195). β values showed additional information, with
hemispheric significance (10.72 ± 3.48 10-3ms-1, 9.69 ±
3.51 10-3ms-1; p<0.037).Discussion/Conclusions
In
this pilot study, we present the first evidence of an MR-based OEF and CVR
technique that requires no contrast. We
have found that MR-PARSE has detectable sensitivity to frequency shifts induced
by transient alterations in de-oxyhemoglobin through the cardiac cycle in ICAD
patients with greater than 50% stenosis.
Furthermore, we have shown that through the use of ICA, transient OEF
and β are significant predictors of hemispheric compromise. Our approach to quantify transient BOLD
fluctuations due to cerebrovascular reactivity represents a new and simple,
non-contrast approach to stratifying patients toward therapies to prevent
stroke.Acknowledgements
R01NS093908; R21EB01792References
[1]
Derdeyn, et al Brain 2002 [2] Yablonskiy et al MRM 1994, [3] Twieg, et al MRM
2002, Menon et al, JCBFM 2012