Magnetic resonance spectroscopic imaging for visualization of the infiltration zone of glioma

In conventional MR imaging, it is often difficult to delineate the heterogeneous structure of gliomas. Proton magnetic resonance spectroscopic imaging ((1)H-MRSI) is a noninvasive tool for investigating the spatial distribution of metabolic changes in brain lesions. The aim of this study was to assess the improvements in delineation of gliomas based on segmentation of metabolic changes measured with (1)H-MRSI.

Twenty patients with gliomas (WHO grade II and III) were examined using a standard (1)H-MRSI sequence. Metabolic maps for choline (Cho), N-acetyl-aspartate (NAA) and Cho/NAA ratios were calculated and segmented based on the assumption of a Gaussian distribution of the Cho/NAA values for normal brain. Areas of hyperintensity on T2-weighted (T2w) MR images were compared with the areas of the segmented tumor on Cho/NAA maps. Stereotactic biopsies were obtained from the MRSI/T2w difference areas.

In all patients, the segmented MRSI tumor areas were greater than the T2w hyperintense areas, on average, by 20% (range 6-34%). In nine patients, biopsy sampling from the MRSI/T2w difference areas showed tumor infiltration ranging from 4-17% (mean 9%) tumor cells, in the areas detected only by MRSI.

Our method for automated segmentation of the lesion-related metabolic changes achieved significantly improved delineation for gliomas compared to routine clinical methods. We demonstrate that this method can improve delineation of tumor borders compared to routine imaging strategies in clinics. Metabolic images of the segmented tumor may thus be helpful for therapeutic planning.