Prediction of glioblastoma multiform response to bevacizumab treatment using multi-parametric MRI

Glioblastoma multiform (GBM) is a highly malignant brain tumor. Bevacizumab is a recent therapy for stopping tumor growth and even shrinking tumor through inhibition of vascular development (angiogenesis). This paper presents a non-invasive approach based on image analysis of multi-parametric magnetic resonance images (MRI) to predict response of GBM to this treatment. The resulting prediction system has potential to be used by physicians to optimize treatment plans of the GBM patients. The proposed method applies signal decomposition and histogram analysis methods to extract statistical features from Gd-enhanced regions of tumor that quantify its microstructural characteristics. MRI studies of 12 patients at multiple time points before and up to four months after treatment are used in this work. Changes in the Gd-enhancement as well as necrosis and edema after treatment are used to evaluate the response. Leave-one-out cross validation method is applied to evaluate prediction quality of the models. Predictive models developed in this work have large regression coefficients (maximum R² = 0.95) indicating their capability to predict response to therapy.