Rationale and Objectives
Metabolite peak boundary definition is an important postprocessing step in proton magnetic resonance spectroscopy ( 1 H-MRS). We compare metabolite ratios calculated using three different postprocessing strategies: software-rendered default peak boundaries, manually adjusted peak boundaries and a curve-fitting program. The first two of these methods are commercially available.
Materials and Methods
A total of 42 spectra acquired on a 1.5-T MR unit using two-dimensional chemical shift proton MR spectroscopy (TR/TE = 1500/144 ms) were analyzed. Choline (Cho), creatine (Cr), and N -acetylaspartate (NAA) relative concentrations were derived and the following metabolite ratios were calculated: Cho/Cr, Cho/NAA, and NAA/Cr. Metabolite concentrations/ratios were calculated using (a) default peak boundaries rendered by commercially available, postprocessing software (Functool 2000, version 2.6.0); (b) manually adjusted peak boundaries by an experienced spectroscopist, using an option offered by the same commercially available software; and (c) an offline in-house curve-fitting program. Measurements obtained with method (c) were considered as the “gold standard.” Paired t -tests comparing default and adjusted metabolite ratios, as well as default and adjusted ratios with their respective curve-fit values were used for statistical analysis.
Results
Significant differences between default and manually adjusted values were found for Cho/Cr ratios <1.5 and for all Cho/NAA ratios. For Cho/Cr ratios <1.5, significant differences between default and curve-fit values were present; this was not the case when comparing manually adjusted and curve-fit values. Default and manually adjusted Cho/NAA ratios were significantly higher than corresponding curve-fit ratios. Manually adjusted values were, however, closer to the curve-fit values. No significant differences were noted between default and adjusted NAA/Cr values; default and manually adjusted ratios were significantly lower than curve-fit ratios.
Conclusion
There can be significant differences in metabolite ratios calculated using default and manually adjusted peak limits in proton MR spectroscopy. Furthermore, Cho/Cr and NAA/Cho adjusted metabolite ratios are closer to curve-fit values, which are considered the most accurate of the three.
In vivo proton MR spectroscopy has become an important tool in the workup of focal brain lesions ( ). Spectral interpretation involves visual analysis of metabolic peaks and metabolite ratio calculations for lesion characterization. Vendor-provided spectral analysis software is commonly used in clinical practice since it offers an efficient and standardized means for data reduction. These vendor-provided software packages may invoke preset boundaries for metabolite peak area integration. In our experience with spectral postprocessing, we have often encountered a visually apparent misalignment of spectral locations using the preset vendor-provided limits relative to the “true” metabolite peaks. This misalignment can present either as a generalized trend or within a subset of voxels in a 2D CSI dataset and could lead to clinically significant alterations in metabolite ratio calculations.
One strategy to approach this apparent discrepancy between preset and “true” metabolite peak limits is to manually adjust the different peak boundaries for each dataset. A more elaborate way of addressing metabolite peak limits is by using a curve-fitting program that ameliorates quantification of partially overlapping peaks ( ). A number of these curve-fitting programs are commercially available. The purpose of our study was to assess changes in calculated metabolite ratios when using preset vendor-provided metabolite peak limits as opposed to manually adjusted peak boundaries and curve-fitting methods.
Materials and methods
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Data Analysis
Get Radiology Tree app to read full this article<
Results
Get Radiology Tree app to read full this article<
Table 1
Summary of Mean Metabolic Ratios Calculated Using the Three Different Methods and Results of Significance Tests for Differences Between Values Rendered by the Different Methods
Metabolite Ratio Range Curve Fit m Mean Ratio Default Limits Mean Ratio Adjusted Limits Mean Ratio Default Versus Adjusted Limits Default Versus Curve-fit Limits Adjusted Versus Curve-fit Limits Cho/Cr <1.3 (n = 9) 1.15 ± 0.13 1.41 ± 0.14 1.14 ± 0.10P < 0.001P = 0.001 NS 1.3−1.5 (n = 11) 1.38 ± 0.06 1.58 ± 0.12 1.26 ± 0.18P < 0.0001P = 0.0001 NS >1.5 (n = 10) 1.78 ± 0.29 1.85 ± 0.35 1.78 ± 0.50 NS NS NS Cho/NAA <0.7 (n = 13) 0.60 ± 0.09 0.99 ± 0.13 0.78 ± 0.10P < 0.0001P < 0.0001P < 0.0001 0.7−1.3 (n = 14) 0.88 ± 0.18 1.23 ± 0.24 1.09 ± 0.29P = 0.001P < 0.0001P = 0.001 >1.3 (n = 14) 1.98 ± 0.49 3.35 ± 1.75 2.77 ± 1.27P = 0.002P = 0.003P = 0.01 NAA/Cr <1.3 (n = 11) 1.15 ± 0.17 0.93 ± 0.26 0.93 ± 0.31 NS_P_ < 0.01P < 0.01 1.3−1.9 (n = 12) 1.65 ± 0.17 1.15 ± 0.33 1.16 ± 0.31 NS_P_ < 0.001P = 0.001 >1.9 (n = 11) 2.43 ± 0.38 1.65 ± 0.19 1.65 ± 0.24 NS_P_ < 0.0001P < 0.0001
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Discussion
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Conclusion
Get Radiology Tree app to read full this article<
References
1. Nelson S.J., Vigneron D.B., Dillon W.P.: Serial evaluation of patients with brain tumors using volume MRI and 3D 1H MRSI. NMR Biomed 1999; 12: pp. 123-138.
2. Nelson S.J., et. al.: In vivo molecular imaging for planning radiation therapy of gliomas: An application of 1H MRSI. J Magn Reson Imaging 2002; 16: pp. 464-476.
3. Weybright P., et. al.: Differentiation of tumor recurrence from treatment related changes using 2D-CSI MR spectroscopy. AJR Am J Roentgenol 2005; 185: pp. 1471-1476.
4. McKnight T.R., et. al.: An automated technique for the quantitative assessment of 3D-MRSI data from patients with glioma. J Magn Reson Imaging 2001; 13: pp. 167-177.
5. Provencher S.W.: Automatic quantitation of localized in vivo 1H spectra with LC Model. NMR Biomed 2001; 14: pp. 260-264.
6. Webb P.G., et. al.: Automated single-voxel proton MRS: technical development and multisite verification. Magn Reson Med 1994; 31: pp. 365-373.
7. Marshall I., et. al.: Reproducibility of metabolite peak areas in 1H MRS of brain. Magn Reson Imaging 1996; 14: pp. 281-292.
8. Vanhamme L.A., van den Boogaart A., Van Huffe S.: Improved method for accurate and efficient quantification of MRS data with use of prior knowledge. J Magn Reson 1997; 129: pp. 35-43.