Click on image to view larger version.
FIG 7. ROC curves generated by varying the nCBV cutoff values used to classify the tumor regions of interest (ROIs) shown in Fig 1 as low grade versus high grade. The left curve (diamond-shaped data points) describes the sensitivity and specificity values for low- versus high-grade astrocytomas only. The right curve (square data points) describes the sensitivity and specificity values for low- versus high-grade astrocytomas and oligodendrogliomas grouped together. Of note, the larger area under the left as compared with the right-shifted curve (proportional to accuracy) reflects the superior discriminatory ability of rCBV imaging in distinguishing low- from high-grade astrocytomas than in distinguishing low- from high-grade oligodendrogliomas. At a critical test cutoff value of nCBV = 1.5 for the right-shifted curve (as per Fig 1), sensitivity and specificity for distinguishing low- versus high-grade glial neoplasms are 0.96 and 0.60, respectively. TPF indicates true-positive fraction; FPF, false-positive fraction.
