This thesis describes the development of prediction techniques for tissue characteristics in brain tumors, using both imaging and tissue information. Magnetic resonance imaging (MRI) serves as an aid in the clinical management of brain tumors, both as a tool for tumor characterization and monitoring, as well as guidance for biopsy sampling. However, the use of conventional MRI can lead to biopsy sampling errors and limited information for tumor analysis. The development of prediction techniques for tissue characteristics in brain tumors could provide valuable, additional information that would assist with both of these tasks. This work describes how such techniques were built using three key steps: data collection and curation, model construction, and model validation. Data collection involved patient screening and recruitment for an IRB-approved, HIPAA-compliant clinical trial protocol. Thirty-one (31) treatment-naïve, adult glioma patients were imaged using an extensive set of MR imaging techniques, from which multiple biopsy targets were specified using conventional and advanced MR image sequences. Biopsy tissue from these target sites was sampled under stereotactic guidance during craniotomy procedures, then stained and analyzed for World Health Organization (WHO) grade, cell proliferation markers, vascularity markers, and cell density. Images were normalized using biological reference regions, registered to the patient’s T2 volume, then sampled at the relevant biopsy sites by propagating a spherical volume of interest (VOI) through all the images. Model construction involved using both linear/logistic regression and random forest (RFT) machine learning techniques to relate 25 imaging parameters to 4 tissue parameters, all obtained from the same biopsy sites. Various statistical tests and analyses were also used to develop different sets of imaging parameters to serve as inputs to the regression and RFT techniques.