Is volume transfer coefficient (K(trans)) related to histologic grade in human gliomas?

Abstract

Purpose: The purpose of this study was to examine the relationship between contrast transfer coefficient [K(trans)] and grade in gliomas.

Material and methods: Median values of K(trans), CBV(T1), and of the 95th percentile (95%) of the distribution (K(trans) [95%] and CBV(T1) [95%]) were calculated in 39 patients with glioma. Group comparisons and post hoc pairwise comparisons were performed. The relationship between variables and grade used Spearman rho and canonical discriminant analysis. The separation of high- from low-grade tumors was separately assessed by using Mann-Whitney U tests and logistic regression. Receiver operator curve analysis was performed for high- and low-grade tumors and grade III and grade IV tumors.

Results: There were significant differences between grades for all variables (P < .001). Pairwise comparisons demonstrated significant differences between grades II and III and II and IV for all variables except K(trans), which did not show significance in the grade II and III comparison, and between III and IV for CBV(T1) and CBV(T1) (95%; P < .01). All variables correlated with grade (P < .01). Discriminant analysis showed independent relation between both CBV(T1) and K(trans) (95%) and grade, and the canonical function produced a total correct classification of 74.4% of cases. Logistic regression analysis for low- versus high-grade tumors showed K(trans) (95%) and CBV(T1) to be independent factors (P < .01 and P < .05).

Conclusion: There are strong independent relationships between both CBV and K(trans) and histologic grade in gliomas. Both measurements show good discriminative power in distinguishing between low- and high-grade tumors with diagnostic sensitivity and specificity >90%.

Fig 1.

Representative K^trans maps from (A) a grade II fibrillary astrocytoma, (B) a grade III anaplastic astrocytoma, and (C) a grade IV glioblastoma multiforme. The white boxes enclose the tumor area in each image. Note that vasculature does not appear in these maps, and K^trans values in normal brain are insignificant and consistent with noise. The K^trans values in the grade II tumor (A) are insignificant corresponding to the lack of enhancement with contrast. The high-grade-defining necrotic core is clearly evident in the middle of the tumor in panel C. The heterogeneity of K^trans is clearly evident in the enhancing tumor portion in panels B and C.

Fig 2.

Representative CBV maps from (A) a grade II fibrillary astrocytoma, (B) a grade III anaplastic astrocytoma, and (C) a grade IV glioblastoma multiforme. The white boxes enclose the tumor area in each image. The normal cerebral vasculature is clearly seen on these maps, particularly the superior sagittal sinus and other major vessels. The grade II tumor in panel A homogeneously shows very low blood volume, which is below the measurement accuracy of the technique. The necrotic core is clearly evident in the middle of the tumor in panel C. The heterogeneity of CBV is clearly evident in the enhancing tumor portion in panels B and C and shows very different distributions to those in Fig 1 (A and B).

Fig 3.

Scattergrams showing the relationships between histologic grade and median values of K^trans, K^trans (95%), CBV, and CBV (95%). Individual cases are indicated by circles, multiple cases are represented by the addition of “petals” to the glyph with the number of petals representing the number of cases. Lines indicate the optimal linear regression fit for the data and the 95th percentile confidence limits for the regression fit for the entire dataset. The correlation between grade and the median values of each of the parametric variables is significant (K^trans, K^trans [95%], CBV, and CBV [95%]; P < .01).

Fig 4.

Scattergram showing the relationship between values of median CBV and K^trans (95%) for all individual cases. The grade II tumors show lower values of both CBV and K^trans (95%). Higher values are seen in grade III and IV tumors, but there is a considerable overlap in these distributions.

Fig 5.

ROC analysis showing effect of using each individual variable or the discriminant function (C1) in differentiating between high- and low-grade tumors. The area under the ROC curve for high- versus low-grade is greatest for the discriminant function (0.993). Within the independent parametric variables the area was highest for K^trans (95%) (0.986), though similarly high values were seen for K^trans and CBV (0.979 and 0.966, respectively). (Areas under the ROC curves are shown in Table 7.)

Fig 6.

ROC analysis for separation of grade III and IV tumors. Areas under the ROC curves are shown in Table 7.