Grading glial tumors with amide proton transfer MR imaging: different analytical approaches

Amide proton transfer (APT) magnetic resonance imaging is gaining attention for its capability for grading glial tumors. Usually, a representative slice is analyzed. Different definitions of tumor areas have been employed in previous studies. We hypothesized that the accuracy of APT imaging for brain tumor grading may depend upon the analytical methodology used, such as selection of regions of interest (ROIs), single or multiple tumor slices, and whether or not there is normalization to the contralateral white matter. This study was approved by the institutional review board, and written informed consent was waived. Twenty-six patients with histologically proven glial tumors underwent preoperative APT imaging with a three-dimensional gradient-echo sequence. Two neuroradiologists independently analyzed APT asymmetry (APTasym) images by placing ROIs on both a single representative slice (RS) and all slices including tumor (i.e. whole tumor: WT). ROIs indicating tumor extent were separately defined on both FLAIR and, if applicable, contrast-enhanced T1-weighted images (CE-T1WI), yielding four mean APTasym values (RS-FLAIR, WT-FLAIR, RS-CE-T1WI, and WT-CE-T1WI). The maximum values were also measured using small ROIs, and their differences among grades were evaluated. Receiver operating characteristic (ROC) curve analysis was also conducted on mean and maximum values. Intra-class correlation coefficients for inter-observer agreement were excellent. Significant differences were observed between high- and low-grade gliomas for all five methods ( P