Meeting Banner
Abstract #3641

Diagnostic Performance of DTI in Differentiating Glioblastomas from Brain Metastases

Sumei Wang1, Sang Joon Kim1, 2, John H. Woo1, Suyash Mohan1, Ruyun Jin3, Matthew R. Voluck1, Ronald L. Wolf1, Donald M. ORourke4, Harish Poptani1, Elias R. Melhem1, 5, Sungheon Kim6

1Radiology, University of Pennsylvania, Philadelphia, PA, United States; 2Radiology, University of Ulsan Asan Medical Center, Seoul, Korea; 3Medical Data Research Center, Providence Health & Services, Portland, OR, United States; 4Neurosurgery, University of Pennsylvania, Philadelphia, PA, United States; 5Radiology, University of Maryland, Baltimore, MD, United States; 6Radiology, New York University School of Medicine, New York, NY, United States

In this study, we investigated the potential of DTI metrics for differentiation tumor types with a substantially larger cohort (n = 222) and also its performance in comparison with two experienced neuroradiologists. 128 glioblastomas and 94 brain metastases were included in this study. Two neuroradiologists independently reviewed the images. Diagnostic performance was evaluated using ROC curve for the two raters and logistic regression model (LRM). Our result indicates that our model is as good as experienced neuroradiologist. Furthermore, it was found that the accuracy of LRM model did not vary as much as those of the raters depending on the selection of the cases.