A

Oli 2 Ast 2 Oli 3 GBM GBM GBM GBM

Oli 2 Ast 2 Oli 3 GBM GBM GBM GBM

Oli 2 Ast 2 Oli 3 GBM GBM GBM GBM

■ CONTRALATERAL

Figure 5. Concentrations (mM) of choline, creatine and NAA in 7 primary, untreated human brain tumors, and in the contralateral hemisphere of each patient. All tumors were subsequently biopsied and evaluated histologically. There are 3 low-grade tumors (Oli 2 - oligodendroglioma grade 2, Oli 3 - oligodendroglioma grade 3, Ast 2 - astrocytoma grade 2) and 4 high grade (GBM - glioblastoma multiforme, grade 4). All tumors have reduced NAA compared to the contralateral hemisphere, while only the high grade tumors have elevated Cho. Cr showed no consistent patterns - variations in Cr in the contralateral hemisphere reflect differences in tumor locations.

However, spectroscopy performed with phased-array coils does increase complexity of the study, especially in terms of spectral analysis and quantitation. Using phased-array coils, each receiver channel generates its own spectrum that has a different sensitivity (depending on coil geometry and voxel location) and phase. In order to produce a single MRS or MRSI dataset with optimum sensitivity, phase-correction and optimal combination of channels must be performed. This can readily be performed using information either in the spectra themselves or through the use of MRI scans to measure the coil sensitivity profiles (45,46).

Local, phased-array coils are intrinsically associated with inhomogenous B1 (RF) fields, and may also exhibit variable sample loading, further complicating quantitation methods, such as the phantom replacement or external reference methods, based on equations (1) and (2). Fortunately, techniques for combining spectra from different channels with uniform sensitivity are available which can correct for variable loading and inhomogenous B1 fields. One example is the SENSE-MRSI method which uses reduced phase-encoding to significantly decrease scan time (45). Multi-channel time-domain MRSI data (b(t,x,y)) can be combined and un-folded using the relationship:

wheres(t,x,y) are combined, uniform spatial sensitivity MRSI data, and A(x,y) is the complex coil sensitivity matrix for each element of the phased-array. A(x,y) can be readily calculated from rapid gradient echo images collected alternatively using the body coil and the phased-array coil.

To calculate metabolite concentrations, a SENSE-MRSI scan must be performed and reconstructed under identical conditions on a calibration phantom of known concentration (in our case a 4 liter sample containing 65 mM NAA). In vivo metabolite concentrations [M] (where M = Choline (Cho), Creatine (Cr) or NAA) can then be calculated using the following expression:

where S is the peak area and LFi and LFNAA are the body coil transmitter load factors for the in vivo and phantom scans, respectively. f(T1,T2) applies a correction factor to account for differences in relaxation times between the phantom and in vivo metabolites. In vivo relaxation times for normal human brain at 3T can either be measured or taken from the literature.

An example dataset using this approach is described below; all scans were performed on a Philips Intera 3.0 Tesla system using a 6-channel phased-array receiver coil. RF pulses were transmitted using the body coil. A 3-slice, spin-echo circularly-encoded 2D-MRSI pulse sequence with water/lipid suppression (47) and OVS, covering from the basal ganglia to the vertex, was collected (TR/TE 2000/144 msec, FOV 230x115 mm, matrix size 32x16, SENSE factor 2, scan time 12 minutes). Prior to MRSI, field homogeneity was optimized using high order shimming. After MRSI, additional rapid gradient echo MRI scans were recorded to calculate the coil sensitivity matrix. The protocol was tested in 5 normal adult subjects (age 34 ±11 years, 4 male). Bilateral metabolite peak areas were measured in 6 representative white and gray matter regions.

Figure 6. An example 3 Tesla SENSE-MRSI data set from a normal human brain, showing representative spectra from white and gray matter locations, as well as metabolic images of choline and N-acetyl aspartate.

Figure 6 shows a representative SENSE-MRSI dataset with 2 of the 6 voxel locations selected for quantitative analysis. Metabolite concentrations for all 6 brain regions analyzed (millimolar, mean ± st.dev.) are given in Table 2. Metabolite concentrations are in reasonable agreement with those from prior MRSI and single-voxel studies that have used conventional head coils (6,29). MRSI with phased-array coils offers improved SNR compared to conventional volume head coils, and allows for shorter scan times using SENSE encoding. However, because of the spatially varying sensitivity (and loading) of the individual receiver coil elements, quantitation is more complex. Fortunately, collection of B1 sensitivity maps and use of SENSE processing allows the reconstruction of uniform sensitivity metabolic images. Since the B1 sensitivity maps are derived relative to the body coil, a global loading correction based on the body coil transmitter gain required for a 90° pulse can be applied. The phantom replacement technique has previously been demonstrated to be a reliable method for quantifying multi-slice MRSI data recorded with homogeneous, quadrature transmit-receive coils (6). The procedure described here allows this method to be extended for use with any type of inhomogeneous coil array, and can be used with either conventional or SENSE-encoding provided that appropriate reconstruction techniques are applied.

0 0

Post a comment