Objective: The objective of our study was to evaluate the diagnostic performance of qualitative 3-T proton MR spectroscopy (1H-MRS) for the characterization of musculoskeletal neoplasms.
Subjects and methods: Proton MRS studies of 74 patients (76 lesions) with a histologically confirmed musculoskeletal neoplasm or neoplasms were prospectively included in this study. All studies were performed using a 3-T MRI scanner. Spectra were analyzed with conventional MRI software provided by the fabricant and with dedicated independent MRS software. Spectra were evaluated visually for the presence or absence of a choline peak at 3.2 ppm. The presence of a choline peak was considered indicative of malignancy. The influences of tumor origin and spectral quality on diagnostic performance were considered.
Results: Diagnostic performance was similar with both software used (κ = 0.97). Qualitative 1H-MRS failed to differentiate benign from malignant bone tumors regardless of the application of quality criteria (best sensitivity and specificity, 50.0% and 61.5%, respectively). Diagnostic performance was better with soft-tissue lesions, but the specificity remained low (best sensitivity and specificity, 82.4% and 64.3%, respectively). The application of spectral quality criteria for the evaluation of soft-tissue tumors led to the exclusion of 35.4% of the evaluated spectra.
Conclusion: Qualitative 3-T 1H-MRS cannot be used to characterize bone tumors. This technique has good sensitivity but poor specificity for the characterization of soft-tissue lesions.
Keywords: characterization; musculoskeletal system; neoplasms; proton MR spectroscopy; qualitative analysis.