Nimg-44. radiomic features of contrast enhanced t1 mri sequences predict survival in primary glioblastoma patients who underwent adjuvant radiation therapy

Abstract PURPOSE Glioblastoma (GBM) is an aggressive primary brain tumor that is monitored through magnetic resonance imaging (MRI). Radiomic analysis offers valuable insight into better understanding the imaging features of these tumors. We tested the hypothesis that changes in radiomic features following radiation treatment predicts overall survival. Radiomic features we were interested in included 2D and 3D shape features, first order features evaluating patterns of voxel intensity within the mask regions, and higher order GLCM features. METHODS Data from 19 patients from our brain bank diagnosed with a GBM at autopsy and underwent radiation therapy were used for this study. T1C tumor masks were created by manually annotating contrast enhancing tumor for patients on pre- and post-treatment MRIs. Pyradiomics was used to extract radiomic features within both pre- and post-radiation tumor masks, including 14 shape features, 18 first order features, and 75 higher order features were compared across timepoints and the difference for each variable was correlated with overall survival. RESULTS Of the tested radiomic features, only the following shape features were found to be correlated with overall survival: Major Axis Length (R = -0.55, p =.02), Maximum 2D Diameter Row (R= -.56, p =.02), Maximum 2D Diameter Slice (R= -.52, p =.03), and Maximum 3D Diameter (R= -.47, p = .04). CONCLUSIONS These results show a decrease in the Major Axis Length, Maximum 2D Diameter, Maximum 2D Diameter, and Maximum 3D diameter of the T1C masks post-radiation correlate with an overall increase in survival. These inverse trending radiomic variables may be useful to assess tumor response to radiation therapy, specifically through assessment of shape features within the contrast-enhancing tumor region. Future research should determine how tumor changes in response to other treatment modalities and/or correlates with other prognostic variables.