Nirl-03 Regression Analysis of Non-Categorical Imaging Parameters in Patients With Brain Metastases

BACKGROUND Estimating survival of patients who have brain metastases (BM) based on risk factors and properties of the metastasis would be useful in clinical practice. Imaging data routinely provides significant variables such as volumes of the tumor, surrounding necrosis, and associated edema. These values could be used to create models that could predict patient survival. Our work applied regression analysis to see how imaging parameters correlated to survival in BM patients. METHODS The open-access dataset Pretreat-MetsToBrain-Masks1,2 from the Yale Department of Radiology and Biomedical Imaging featuring 200 BM patients was accessed through the Cancer Imaging Archive. We utilized the Analysis TookPak in Microsoft Excel to perform regression analysis of all non-categorical variables with respect to survival. We performed multivariate regression between survival, edema volume, and necrosis volume. Patients with missing data were omitted from the regression analysis. RESULTS There is low correlation between survival and the individual parameters of necrosis ratio (R = 0.12), edema ratio (R = 0.06), necrosis volume (R = 0.02), edema volume (R = 0.06), enhancing tumor (ET) volume (R = 0.18), number of edema lesions (R = 0.10), number of necrotic lesions (R = 0.07), or number of ET lesions (R = 0.05). However, the relationship between enhancing tumor volume and survival does show significance (p = 0.01). The multivariate regression between survival, edema volume, and necrosis volume showed low correlation (0.05) and no statistical significance. CONCLUSION There is a significant need to estimate survival in BM patients. However, based on our analysis, we did not find utility in the assessed imaging parameters to estimate survival via a linear model. Still, ET volume was shown to have a significant relationship with survival, and a different model needs to be used to evaluate that relationship.