NIRA-02 REAL-WORLD ANALYSIS OF THE PREDICTIVE VALUE OF DCE MR PERFUSION AND MULTIDISCIPLINARY BRAIN METASTASIS TUMOR BOARD REVIEW IN DISTINGUISHING PATHOLOGICALLY-CONFIRMED RADIATION NECROSIS VERSUS POST-RADIATION RECURRENCE IN MELANOMA

Abstract INTRODUCTION Distinguishing recurrent brain metastases (rBrM) from radiation necrosis (RN) is a diagnostic challenge due to their potentially similar appearance, even on advanced imaging. Approximately 15% of SRS-treated brain metastases (BrM) progress within 1 year of radiotherapy and MR perfusion imaging is used as a noninvasive indicator for recurrent tumor. However, in melanoma patients, its specificity in identifying recurrence is unclear, due to melanocytic and hemorrhagic radiographic confounders. We hypothesize that recurrent melanoma BrM exhibit atypically low plasma volume (Vp), increasing the risk of misdiagnosis. Furthermore, we hypothesize that prospective multidisciplinary tumor board (MTB) consensus offers improved prediction of pathologic diagnosis over Vp alone. METHODS Patients with post-SRS melanoma BrM that underwent surgical resection, with pre-resection MR DCE perfusion imaging, between 1/2014-3/2025 were identified. Lesion classifications of Vp(high) or Vp(low), as assigned by blinded neuroradiologists, were compared to neuropathologic diagnoses. Lesions were categorized as rBrM, RN, or mixed for quantitative analysis which is forthcoming. The predictive value of MTB conclusions was determined. RESULTS 73 lesions in 67 melanoma patients were included: 53 (73%) pathologically-confirmed rBrM and 20 (27%) RN. Qualitative analysis demonstrated that of RN cases, 6 (30%) were Vp(low), predicting RN; however, 14 (70%) were Vp(high), which ordinarily predicts rBrM. Meanwhile, 21 (40%) of rBrM cases were Vp(low) and 32 (60%) were Vp(high). The positive predictive value (PPV) of MR perfusion for rBrM was 70%, while the negative predictive value (NPV) was 22%. For cases with preoperative MTB predictions, there was an improvement in predictive values, with PPV of 80% and NPV of 60%. CONCLUSION MR perfusion has limited predictive value in distinguishing rBrM from RN in post-SRS melanoma BrM. Multidisciplinary consensus, however, improves diagnostic accuracy over MR perfusion alone, highlighting the value-added role of such review in guiding care of complex patients.