Moffat, et al. utilized functional diffusion map imaging biomarker and concluded that chemotherapy dose was correlated with this biomarker and the dose itself was also correlated with the response. Lemaire, et al. examined tumor treatment in rats and reached the conclusion that there were some relationships between pre-treatment diffusion weighted parameters and the tumor size several days after the therapy. Baurle, et al. investigated the predictability of the response in patients with breast cancer bone metastasis and showed that the change in the lesion size can be assessed much earlier via the DCE-MRI biomarkers. Swanson, et al. developed a model for computing the rate of change in the glioma cell concentration and for estimating the patients’ survival, mainly based on two biological factors. Development of a prediction system requires at least two Fulvestrant moa series of images acquired from a number of patients to specify some measure of response. Additionally, using serial images, changes of specific biological and imaging parameters may be traced and their relationship with treatment and time may be investigated. Therefore, many studies have focused on this aspect of medical imaging. There are several sources of error and variance in these images that should be carefully considered in their analysis. The purpose of this work is to establish a relationship between multi-parametric MRI, including T1-weighted pre-Gd, T1-weighted post-Gd, T2-weighted, and Fluid attenuated inversion recovery images acquired pretreatment, and the reduction in the Gd-enhanced volume due to bevacizumab treatment. The differences among the Gd-enhanced regions of different patients in terms of their homogeneity and brightness has motivated us to extract their characteristics and features to stratify responders from non-responders and develop a predictive model for the level of response. In addition, analysis of the data acquired from the patients in several consecutive imaging series is performed to see how the patients’ conditions are affected by the therapy and how the tumor characteristics are influenced during the treatment time interval. To the best of our knowledge, this work is the first study that uses multi-parametric structural MRI to predict the response to therapy. To define the Gd-enhanced area, the T1-post image was divided by the T1-weighted image pixel by pixel and the result theresholded. This method requires that the two images have similar brightness. To this end, a normalization step was applied to the images by selecting an ROI in the unaffected WM on the T1- pre image and its corresponding region in the T1-post image. Then, the average intensities of the pixels in this region of the two images were calculated and the relative gain of the two images was obtained by dividing their average intensities. The gain was used for the normalization of the images. The process of ROI definition was performed for the edema and necrosis as well. For this aim, a simple thresholding was applied to the FLAIR and T1-post images to extract edema and necrosis, respectively. To treat all of the ROIs equally, an identical threshold should be used for all of the images from the same modality. Therefore, the adverse effect of the intensity gain in some images was eliminated by normalization of the intensities. For example, to extract the ROI of edema, the edema in a sample FLAIR slice was first segmented manually.