Brain tumor classification by proton MR spectroscopy: comparison of diagnostic accuracy at short and long TE

Abstract

Background and purpose: Different TE can be used for obtaining MR spectra of brain tumors. The purpose of this study was to determine the influence of the TE used in brain tumor classification by comparing the performance of spectra obtained at two different TE (30 ms and 136 ms).

Methods: One hundred fifty-one studies of patients with brain tumors (37 meningiomas, 12 low grade astrocytomas, 16 anaplastic astrocytomas, 54 glioblastomas, and 32 metastases) were retrospectively selected from a series of 378 consecutive examinations of brain masses. Single voxel proton MR spectroscopy at TE 30 ms and 136 ms was performed with point-resolved spectroscopy in all cases. Fitted areas of nine resonances of interest were normalized to water. Tumors were classified into four groups (meningioma, low grade astrocytoma, anaplastic astrocytoma, and glioblastoma-metastases) by means of linear discriminant analysis. The performance of linear discriminant analysis at each TE was assessed by using the leave-one-out method.

Results: Tumor classification was slightly better at short TE (123 [81%] of 151 cases correctly classified) than at long TE (118 [78%] of 151 cases correctly classified). Meningioma was the only group that showed higher sensitivity and specificity at long TE. Improved results were obtained when both TE were considered simultaneously: the suggested diagnosis was correct in 105 (94%) of 112 cases when both TE agreed, whereas the correct diagnosis was suggested by at least one TE in 136 (90%) of 151 cases.

Conclusion: Short TE provides slightly better tumor classification, and results improve when both TE are considered simultaneously. Meningioma was the only tumor group in which long TE performed better than short TE.

Fig 1.

Mean spectra at short TE (TE = 30) of the five tumor types included in the study.

Fig 2.

Mean spectra at long TE (TE = 136) of the five tumor types included in the study.

Fig 3.

Box plots show the distribution of some relevant resonances. The horizontal line is the median, the ends of the boxes are the upper and lower quartiles, and the vertical lines show the full range of values in the data. The extreme points (*) are outliers. A, LIP13 at short TE. B, LIP13 at long TE. C, Ala at short TE. D, Ala at long TE. E, CR at short TE. F, CR at long TE. G, CHO at short TE. H, CHO at long TE. I, Gly/MI at short TE. J, Gly/MI at long TE.

Fig 4.

Scatter plots show the distribution of the complete set of cases at short TE (A) and long TE (B) by using the two first discriminant functions obtained for linear discriminant analysis (x axis, value obtained with the first discriminant function; y axis, value obtained with the second discriminant function). Open circle indicates meningiomas; open square, low-grade astrocytomas; open diamond, anaplastic astrocytomas; open triangle, glioblastomas-metastases. The large black symbols show the centroid for every tumor group. With this method, new cases with unknown diagnosis are classified according to their distance to the centroids.