In this dissertation, we propose the first, to the best of our knowledge, PCA based algorithm to noninvasively recognize and classify different temporal stages of brain tumors given a large time series of MRI images. We propose an algorithm that addresses the challenging task of classifying stage of tumor over period of time while the tumor is being treated with VB-111 virotherapy. Our approach treats stage tumor recognition as a two-dimensional recognition problem. Detecting the stage of the tumor is a crucial prognosis factor for predicting the progression of cancer and patient survival. Accurate identification of brain tumor in longitudinal MRI is important for therapy response assessment. We propose a new framework to detect and classify temporal longitudinal MRI with high accuracy rates. A sensitivity rate of 98.7%, 95.8% and 94.01% for stage 1, 2 and 3 respectively are reported. These results agree with the ground truth of MRI scans.

In the second section of this dissertation, we propose a novel mathematical model that describes the complex interaction between tumor cells, the immune system and the novel anti-angiogenic virotherapeutic VB-111. This is the first agent based on a transcription-controlled gene therapy that selectively targets tumor endothelial cells. VB-111 is an engineered adenovirus which has been previously shown to have antitumor properties in vitro and in vivo. The goal of our model is to confirm and capture the decay and stabilization of tumor cells by VB-111 monotherapy. The model consists of a system of nonlinear ordinary differential equations describing tumor cells, effector cells, cytokine tumor necrosis factor alpha (TNF-α) and the therapeutic protein Fas-c. Through numerical simulations and stability analysis, we compare the dynamics of two cases: with and without therapy. We show that our mathematical model indeed confirms the efficacy of VB-111 in targeting endothelial tumor cells.