1. Voxel Location Dependence of Brain Metabolites as Determined by Magnetic Resonance Spectroscopy L.Ewell, A. Bhullar and B. Stea—Department of Radiation Oncology, University of Arizona American Association of Physicists in Medicine, Philadelphia, Pennsylvania, 7/18/2010 Abstract Multivoxel Magnetic Resonance Spectroscopy (MRS) has the potential to aid the diagnosis of brain disease, by providing information on the amount and location of metabolites such as Choline (Cho), Creatin (Cr) and N-Acetyl Aspartate (NAA). The absolute metabolite levels are determined by fitting the respective peaks, and determining the Area Under the Curve (AUC). These levels show substantial variation with location, as determined by phantom scans, as well as by healthy volunteers and imaging protocol patients. In an attempt to better understand this variation, the spatial functional dependence of metabolite levels is determined via polynomial curve fitting, and evaluated by R 2 (Pearson product). A second degree polynomial describes the phantom metabolite variation well (R 2 =0.97). For healthy volunteers, the spatial variation is also described well by a second degree polynomial, albeit with a lower quadratic coefficient. For protocol patients and others, the presence of lactate/lipid complicates the functional dependence, especially for the NAA peak which is located closest the lactate/lipid. These relationships are explored in order to detect systematic trends that may assist in diagnosis. Introduction While the metabolite information afforded by MRS shows great potential, the clinical benefits remain elusive. For example, the Center for Medicare and Medicaid Services (CMS) recently reaffirmed that it would not cover MRS for routine patient care 1. An accurate map of metabolite level spatial/anatomical dependence would help to standardize diagnostic interpretations across different patients, different diseases and different institutions. To this end, phantom MRS spectra, along with healthy volunteers and protocol patients have been studied in an attempt to map out the spatial metabolite level dependence. Methods and Materials Point Resolved SpectroScopy (PRESS) multi-voxel spectroscopy with a 7x7 grid (2x2cm individual voxel size) and a one cm slice thickness were used for all MRS scans. Phantom metabolite concentrations were chosen to be approximately equal to a healthy human brain: 3.0, 10.0 and 12.5mM for Cho, Cr and NAA respectively. A picture of the MRS voxel grid superimposed on an MRI scan for the phantom and a healthy volunteer are displayed in Figure1. For each voxel, individual peaks are fit using a Gaussian function. The raw spectra are extracted from the scans using Functool © (General Electric). Internal Review Board (IRB) approval for human subjects was obtained. Figure 1: MRS Voxel grid. 1A) Phantom. 1B) Healthy volunteer. 1A 1B Results Phantom - Absolute phantom metabolite levels (AUC) are plotted as a function of voxel number in both the left - right (vs. column index) and anterior - posterior (vs. row index) direction. Figure 2 shows a typical Gaussian fit (2A) and a series of curves representing the absolute Cho concentration as a function of voxel row index (anterior - posterior direction) for different voxel columns(2B). A second degree polynomial was fit to the set of six curves: three different metabolites x two different directions. Table 1 lists the R 2 (Pearson product) values of the fits, along with the standard deviation. As can be seen in this Table, the left-right metabolite concentration variation deviates from the second order polynomial slightly more (lower R 2 ) than in the anterior - posterior direction. In the left right direction, the NAA concentration deviates more than the Cho or Cr. DirectionMetaboliteAverage R 2 Standard Dev. Left-RightCho0.970.01 Left-RightCr0.970.01 Left-RightNAA0.930.08 Anterior-PosteriorCho0.970.03 Anterior-PosteriorCr0.980.02 Anterior-PosteriorNAA0.970.01 Table 1: Fit values to quadratic function for phantom metabolites Figure 2: 2A) Gaussian fit to metabolites. 2B) Phantom voxel metabolite levels. 2A 2B Healthy Volunteer – As depicted in Figure 1B, the central 20 voxels (inside red box) are analyzed. In Figure 3, the left-right (sagittal) metabolite variation is shown, averaged over four rows. As can be seen, the AUC variation forms an M, with the most medial voxel having a relatively low metabolite level, followed by a relative maximum on either side in the mid-lateral cerebrum, with relative minima in the most lateral voxels. The coronal (anterior-posterior) variation of the healthy volunteer shows a variation similar to the quadratic one exhibited by the phantom, but with a smaller x 2 coefficient, as discussed below. Protocol Patient – An imaging protocol was initiated to study the ability of MRS to assist the differential diagnosis of radiation necrosis and recurrent disease in patients treated with radiation for gliomas (see http://www.u.arizona.edu/~lewell/protocol/index.html ). It has been known for some time that disease of this nature can lead to increased levels of lactate and/or lipid 2. In Figure 4, the MRS spectra of a protocol patient is displayed. As can be seen in this spectra, a lactate/lipid peak is visible to the right of the NAA peak. http://www.u.arizona.edu/~lewell/protocol/index.html Figure 4: Protocol Patient Spectra Fit lactate/lipid Figure 5A) Coronal NAA Variation. 5B) Coronal Cho Variation MediumMetaboliteQuad. Coef.R2R2 MediumMetaboliteQuad. Coef.R2R2 PhantomCho-10.10.996PhantomNAA-18.90.995 Health Vol.Cho-8.50.996Health Vol.NAA-8.90.840 PatientCho-4.10.999PatientNAA-2.70.122 Table 2: Cho and NAA Fit Parameters Conclusion MRS continues to be a challenging method by which to diagnose brain disease. By determining the functional fit of the spatial dependence of metabolite levels in phantom, healthy volunteer and protocol patients, it is hoped that a higher level of standardization can be achieved, thereby facilitating more widespread use and enabling an increase in the quality of life for this patient population. Acknowledgement This work was supported by the Arizona Biomedical Research Commission – Grant number 0725. References 1. Hollingworth W., et al, “A Systematic Literature Review of Magnetic Resonance Spectroscopy for the Characterization of Brain,” AJNR 2006; Vol. 27: 1404-1411. 2. Imamora K., “Proton MR Spectroscopy of the Brain with a Focus on Chemical Issues,”, Mag. Res. in Med. Sciences, Vol. 2, No. 3, p.117-132, 2003. Discussion As can be seen in Table 1, the left-right (sagittal) metabolite variation in the phantom is in general well fit by second order polynomials, with values of R 2 close to one. However, as can be seen in Figure 3, the sagittal variation in the human brain is more well represented by an M, with the most medial voxel having a relatively low amount of brain metabolites. This shape can likely be explained by brain anatomy: The hemispheres meet in this area, and there are also the ventricles which are filled with cerebral spinal fluid (CSF) and may have low amounts of brain metabolites. Figure 5 displays the coronal (anterior-posterior) variation of both NAA (5A) and Cho (5B) for the phantom, healthy volunteer and protocol patient. Table 2 displays the quadratic fit parameters of these data. As can be seen, the quadratic coefficients of the fit to the phantom data have a higher absolute value than do the fits of the human brains: i.e., the parabolic fits to the actual brains are more shallow. For Cho variation, all three data sets are fit well by a second order polynomial, with R 2 values very close to one. On the other hand, the NAA data for the protocol patient has a poor quadratic fit, with a low (0.122) value of R 2. This difference between the healthy patient and the protocol patient can likely be explained by the presence of lactate/lipid in MRS spectra of the patient. The fact the lactate/lipid peak is located close to the NAA (shift lactate/lipid ≈1.6ppm, shift NAA ≈2.0ppm) means that in the protocol patient, the NAA peak may be affected more by the presence of the lactate/lipid, than either the Cho or Cr. Figure 3: Sagittal variation in healthy volunteer