In vivo proton MR spectroscopy of primary and nodal nasopharyngeal carcinoma

Abstract

Background and purpose: The aim of this study was to determine the feasibility of performing in vivo proton ((1)H) MR spectroscopy of nasopharyngeal carcinoma (NPC) and to document the (1)H spectrum of this cancer.

Methods: Twenty-seven patients with NPC lesions >1 cm(3) underwent localized (1)H MR spectroscopy performed at 1.5 T. Water-suppressed spectra from both primary tumors (nine cases) and metastatic nodes (18 cases) were obtained at TE 136 and 272. Spectra were analyzed in the time domain by using a nonlinear least squares fitting algorithm with incorporation of previous knowledge. Choline (Cho)/creatine (Cr) ratios for primary NPC and metastatic nodes were calculated and compared. Spectra from normal neck muscle of five volunteers were acquired as control data.

Results: (1)H MR spectroscopy was successfully obtained in seven (78%) of nine primary tumors and 16 (89%) of 18 metastatic nodes. Intense lipid signals in the range of 0.89 to 2.02 ppm were observed in 95% of spectra at TE 136 and 91% of spectra at TE 272. At TE 136, Cho/Cr for metastatic nodes (5.3 +/- 1.6) was significantly higher than the ratio for primary (2.6 +/- 0.5) NPC lesions (P =.02). Cho/Cr ratios for NPC lesions were higher than those for normal neck muscles, for which values ranged from 0 to 0.97 and 0 to 1.1 at TE 136 and 272, respectively.

Conclusion: (1)H MR spectroscopy is a feasible technique for the evaluation of NPC tumors >1 cm(3). Cho/Cr ratios for the lesions were high compared with those for normal neck muscle.

Fig 1.

Images of patient 1. A, Axial view fat-suppressed T2-weighted MR image shows the position of the volume of interest for spectroscopy represented by the white square placed within the primary NPC tumor. B, Spectrum acquired at ¹H MR spectroscopy with TE 136 from the primary NPC tumor shown in A. The nominal voxel volume was 3.9 cm³. Prominent peaks detected were Cho (3.2 ppm), Cr (3.03 ppm), and lipids (1.30 ppm). C, Lipids were removed from the original spectrum shown in B with the use of Hankel-Lanczos singular value decomposition filtering before spectral fitting was performed (bottom trace). The fitted spectrum performed in the time domain is shown by the middle trace, and the top trace represents the residual signal intensity.

Fig 2.

Images of patient 20 with NPC. A, Coronal view T2-weighted MR image shows the position of the volume of interest for spectroscopy represented by the white rectangle placed within the metastatic lymph node. B, Spectrum acquired at ¹H MR spectroscopy with TE 136 from the metastatic node. The nominal voxel volume was 2 cm³. Both Cho (3.2 ppm) and methylene lipid (1.30 ppm) peaks were present, but Cr (3.02 ppm) was not clearly seen. C, Lipids were removed from the original spectrum shown in B with the use of Hankel-Lanczos singular value decomposition filtering before spectral fitting was performed (bottom trace). The fitted spectrum performed in the time domain is shown by the middle trace. Only the intense Cho peak was successfully estimated, whereas the Cr peak was not found. The top trace represents the residual signal intensity.