Contrast Leakage Patterns from Dynamic Susceptibility Contrast Perfusion MRI in the Grading of Primary Pediatric Brain Tumors

Advances in Knowledge

A striking finding in our data was 100% positive predictive value (25 of 25) for low-grade tumors if a T1-dominant leakage pattern was identified. The converse is also true in that the lack of a T1 leakage pattern (T2* or return to baseline) resulted in a 100% negative predictive value for high-grade tumors. T1-dominant leakage patterns were also sensitive and specific for either a pilocytic astrocytoma or pilomyxoid astrocytoma. In addition, good interobserver agreement for identifying a T1-dominant leakage pattern was found, indicating potential clinical utility for grading pediatric brain tumors with TSITC from DSC perfusion.

Little prior data exist in the literature regarding the use of DSC perfusion MR imaging and signal-intensity curves in the evaluation of pediatric brain tumors. Similar to our findings, Grand et al¹⁸ reported a series of 9 pilocytic astrocytomas evaluated with DSC perfusion MR imaging, in which all 9 tumors demonstrated an overshooting of the baseline or a T1-dominant contrast leakage pattern. Cha et al reported that pilocytic astrocytomas have >70% signal recovery, while medulloblastomas have <50% signal recovery, which is consistent with our findings.¹⁹

Before our study, a T1-dominant leakage pattern has most notably been demonstrated in primary CNS lymphoma and has been shown to be useful in distinguishing lymphoma from glioblastoma multiforme and CNS metastases.^9⇓–11 Despite the lack of neoangiogenesis in CNS lymphoma, both glioblastoma and lymphoma have elevated rCBV but a significant difference in the high percent signal recovery for lymphomas. These features have been attributed to the characteristic perivascular invasion of lymphoma cells, leading to disruption of the basement membrane and resulting vascular permeability. The hypercellularity of lymphoma with a small extravascular extracellular space has been suggested as a contributing factor, though the exact mechanism for T1-dominant leakage effects in lymphoma is not well-understood.⁹ This finding contradicts the findings of Cha et al¹⁷ in that the lack of a blood-brain barrier in metastatic disease, with resulting vascular permeability, results in a low percent recovery or a T2*-dominant effect rather than the T1-dominant effect in CNS lymphoma.^9,17 Cha et al¹⁷ used a flip angle of 35° compared with 80° for Mangla et al,⁹ which may contribute to the differing TSITC results. Biologic factors may include differing rates of contrast leakage between CNS lymphoma and metastatic disease with a faster accumulation and concentration of extravascular contrast in metastatic disease leading to T2*-dominant effects. The mechanism underlying the observed postbolus T1-dominant effects in low-grade pediatric brain tumors is not known but is also likely related to the balance between T1-shortening and T2* effects of paramagnetic contrast medium that has leaked into the extravascular extracellular space. Other factors affecting the T1–T2* balance may include characteristics of the extravascular space such as cellular attenuation, water content, and other molecular constituents, in addition to hemodynamic flow issues and capillary permeability.

Pilocytic astrocytomas and pilomyxoid astrocytomas account for 21 of the 25 tumors in our series demonstrating postbolus T1-dominant leakage. Pilocytic astrocytomas are described as having a biphasic histology with areas of loose glial tissue interposed with compacted piloid tissue composed of attenuated hypercellular sheets of elongated bipolar cells.²⁰ Distinguishing features of pilomyxoid astrocytomas include monomorphous piloid cells in a myxoid background with an angiocentric pattern, similar to perivascular pseudorosettes in ependymomas. While pilomyxoid astrocytomas tend to occur in younger children in the hypothalamic/chiasmatic pathway, with more aggressive behavior and a higher propensity for leptomeningeal spread, pilocytic and pilomyxoid astrocytomas have large morphologic and imaging overlap. There have been reports of recurrent pilomyxoid tumors demonstrating pilocytic features after several years, with some authors suggesting that pilomyxoid astrocytoma may be an infantile form versus an extreme subtype of pilocytic astrocytoma.^21⇓–23 Because of this overlap and the lack of distinguishing imaging characteristics between the 2 tumors, we calculated diagnostic accuracy with both tumors rather than separately. Vascular proliferation can be seen in the solid component of both tumors but with more mature endothelial cells in a single layer compared with glioblastomas.^24,25 These single endothelial layers with open tight junctions and fenestrae and hyaline degeneration around vessels have been postulated to allow contrast medium extravasation and therefore vascular permeability.¹⁸ Similar to CNS lymphomas, this level of increased vascular permeability likely plays a partial but crucial role in the T1-dominant leakage effects seen in both tumors. Further research between different pediatric tumors by using T1 signal–based dynamic contrast-enhanced MR imaging may be helpful in elucidating differences in vascular permeability.

Effect of Dexamethasone on Leakage Patterns

We could not control for the administration of dexamethasone in pediatric patients with newly diagnosed brain tumors because corticosteroids are a first-line emergent therapy for controlling the cerebral swelling and the resultant mass effect from mass-occupying brain tumors. In a previous work on the same patient population, there was no significance between patients with and without dexamethasone treatment and rCBV measurements.² However, given the known rapid effects of dexamethasone in decreasing capillary permeability and reducing vasogenic edema,^26,27 the significance between patient populations is not surprising. However, only our low-grade tumor group was significant, and there was a larger proportion of T1-dominant and T2*-dominant leakage patterns within the dexamethasone group. This result seems counterintuitive because high-grade tumors typically have greater vasogenic edema and mass effect, and decreasing capillary permeability should trend toward the return to baseline pattern because this pattern would best represent no leakage. In reality, multiple physiologic parameters likely affect the leakage pattern as indicated by our results. Despite the significance of dexamethasone on the low-grade tumor group, the 100% specificity of a T1-dominant leakage pattern for low-grade pediatric tumors and the high sensitivity and specificity for piloid astrocytomas remain unchanged for both groups with or without dexamethasone.

Implications for Patient Care

Despite uncertainty in the exact underlying mechanism, a T1-dominant contrast leakage pattern on the signal-intensity curve obtained in the region of highest tumor rCBV is empirically highly predictive for low-grade brain tumor in the pediatric population, while a T2*-dominant or baseline pattern has high sensitivity for a high-grade tumor. In addition to rCBV data,² TSITC from DSC perfusion potentially offer additional information not otherwise apparent with routine anatomic MR imaging sequences. Recognition of a T1-dominant leakage pattern may improve accuracy in preoperatively predicting the presence of a low-grade pediatric brain tumor. While experienced pediatric neuroradiologists can recognize low-versus-high-grade tumors in most cases, recognition of this pattern of perfusion leakage may be very helpful in the minority of cases in which low-grade pilocytic or pilomyxoid astrocytomas become large and heterogeneous, mimicking higher grade neoplasms and, conversely, when high-grade astrocytomas have benign imaging features (Figs 1 and 2). Furthermore, when complete surgical resection is not possible, knowing the likelihood of a high-grade or low-grade tumor can guide how aggressive a neurosurgeon should be in debulking the tumor, at the risk of morbidity and mortality.

Limitations

The neuroradiologists evaluating the perfusion data were not blinded to the anatomic MR imaging sequences, some of which were highly suggestive of the tumor diagnosis. This scenario was unavoidable because the anatomic images provide information integral to the assessment of the perfusion images (eg, location of tumor, location of major vessels, and so forth). This limitation is reasonable however because it reflects the process in which MR perfusion images are evaluated in clinical practice. As noted in the discussion, a T1-dominant contrast leakage pattern has been described in CNS lymphoma, and our series of 63 pediatric brain tumors included no cases of CNS lymphoma. While this represents a potential pitfall in using a T1-dominant leakage pattern to classify pediatric brain tumors as low grade, CNS lymphoma is rarely encountered in the pediatric population. Contrast medium infusion rate and peripheral intravenous catheter gauge are related issues that likely affect the quality of the perfusion study and are recognized challenges in performing DSC perfusion MR imaging on pediatric patients.²⁸ However, only 1 study of the 65 identified by the search criteria was excluded for poor contrast medium bolus, indicating that high-quality DSC perfusion MR imaging can be performed in pediatric patients with brain tumors. Differences in TR and TE parameters corresponded to those in studies obtained on different 1.5T and 3T scanners and theoretically have some effect on signal intensity and recovery/leakage patterns. However, our analysis demonstrates no significant differences in our results attributable to these different field strengths or parameters. There are no standardized criteria in the literature for T1 leak/T2* leak or return to baseline. We arbitrarily chose the ±10% range for the curve classification as a means of separation; consequently, curve classification could be altered if different thresholds are used. Similarly, as previously described, larger differences in technical parameters (flip angle, TE, TR) may result in greater differences in curve classification.