NIMG-61: QUANTITATIVE MRI MORPHOLOGIC CHARACTERISTICS AND QUANTITATIVE HISTOLOGIC PROFILES IN SURGICALLY PROVEN RADIATION NECROSIS VERSUS RECURRENT BRAIN TUMOR

BACKGROUND: Distinguishing radiation necrosis (RTN) from recurrent brain tumor (RBT) by standard MRI morphology poses significant challenges. In addition, histologic features associated with RTN and RBT have not been quantified, leading to diagnostic imprecision at lesion resection. We quantitated morphologic MRI characteristics and histologic features in RTN and RBT in irradiated brain tumor patients undergoing lesion resection of imaging progression. METHODS: Retrospective review of irradiated brain tumor patients with imaging progression identified at lesion resection to be predominant RTN (>80% radiation necrosis) or predominant RBT ( >80% active tumor). Two neuroradiologists blinded to diagnosis quantitated (0, < 25%, 25-50%, 50-75%, and > 75%) presurgical MRI features. A profile of histologic findings was similarly scored by one neuropathologist blinded to diagnosis. RESULTS: Study group includes 54 cases for MRI review (30 RBT, 20 RTN) and 38 for histology review (20 RBT and 18 RTN). In a multivariate model of quantitated imaging features, probabilities that RTN and RBT can be estimated in glioma patients are 71% for RBT, 65% for RTN via ROC analysis. T1 hyperintensity, necrosis, spreading wavefront and "Swiss cheese/soap bubble" appearance are moderately correlated with RTN. In brain metastasis, 89% of RTN and 70% of RBT cases are correctly classified. In pathology specimens, significant differences in severity of zonal geographic necrosis in glioma (p = 0.002) and metastasis (p = 0.012) and severity of vascular hyalinization in glioma (p= 0.44) were identified. Fibrinoid and coagulative necrosis and vessel wall thickening were not significantly different in RTN versus RBT. CONCLUSIONS: Quantitative assessment of morphologic MRI abnormalities can assist in distinguishing RTN from RBT, especially in brain metastasis. Quantitative assessment of histologic features identified features significantly correlated with RTN versus RBT. A larger sample size is needed to confirm these findings and is ongoing. Grant support: UL1 RR024989 from NCRR/NIH