Program Information
Assessment of Concurrent Stereotactic Radiosurgery and Bevacizumab Treatment of Recurrent Malignant Gliomas Using Multi-Modality MRI Imaging and Radiomics Analysis
W Sun1,2*, C Wang1 , J Kirkpatrick1 , Z Chang1 , F Yin1 , (1) Duke University Medical Center, Durham, NC, (2) Institute of Information Science and Engineering, Jinan, Shandong,
Presentations
SU-K-708-14 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 708
Purpose: To assess the outcome of recurrent malignant gliomas (MG) patients treated by concurrent stereotactic radiosurgery (SRS) and Bevacizumab (BVZ) using multi-modality MRI imaging and radiomics analysis.
Methods: Thirteen patients with recurrent MG received MRI scans before SRS, 1-week and 2-month after SRS were retrospectively studied. The imaging protocol included T1- and T2-weighted, dynamic contrast-enhanced MRI, and diffusion weighted MRI. Radiomics analysis was performed to extract imaging features from both clinical target volume (CTV) and normal tissue volume receiving more than 12 Gy excluding CTV, V12Gy. Statistical tests (Wilcoxon signed rank tests, p-value<0.01) were performed with Bonferroni correction to assess the changes of radiomics features 1-week and 2-month after SRS. The changes between different WHO grades were evaluated. Correlation tests were used to examine the relationships between the changes of radiomics features and patient survival time. Selected features were used to predict the patient survival time after treatment using Support Vector Regression (SVR) with cross validation.
Results: 17/123 radiomics features from CTV and 33/31 from V12Gy showed significant changes 1-week/2-month after SRS, respectively. Between WHO Grade 3/4 patient groups, 1/14 radiomics features from CTV and 4/4 from V12Gy showed significant differences 1-week/2-month after SRS. The correlation coefficients between the changes of 7/3 radiomics features 1-week/2-month after SRS and the patient survival time were larger than 0.8. Using 2 selected radiomics features from T1w scans and 2 from DCE parametric maps, 8 out of 12 patients’ survival time could be accurately predicted (Mean-Absolute-Error=1.47 months, Root-Mean-Square-Error=2.10 months).
Conclusion: Preliminary results demonstrate the correlated relationship between the changes of radiomics features in both CTV and V12Gy from both functional and anatomical MRI images, especially from T2w, and early treatment response. The results also suggested the potential of using radiomics features in treatment outcome prediction for individualized treatment regime optimization.
Contact Email: