Brain tumor growth simulation: model validation through uncertainty quantification

Brain tumors are one of the main worldwide causes of mortality and morbidity and a critical issue in health risk. Tumor growth prediction is a proper method for better understanding the phenomena and choosing the appropriate therapy for patients. Since tumors' physiological and morphological properties vary significantly in different individuals, using patient specific data is valuable for modelling tumor growth in staging and personalized-therapy planning. However, the validity of the models should be evaluated for their precision assessment based on the decision criteria. There are different sources of uncertainties affecting model prediction accuracy and decision making for the therapy. In this paper, an image-based tumor growth model is evaluated by taking into account uncertainties in the model parameters. The proposed reaction-diffusion model integrates cancerous cell proliferation and invasion through reaction and diffusion terms, respectively. Uncertainties in diffusion and proliferation coefficients were analyzed through Monte Carlo simulation. The time needed for tumor to grow to its fatal size was estimated through numerical solution of the model. Comparison of the predicted time distribution with and without considering uncertainties in model parameters shows a decrease in dispersity of predicted data that highlights the importance of uncertainty. Also, the wide range for survival time shows the importance of choosing proper parameters in order to enhance model accuracy. The recommendations were made for increasing the validity of the tumor growth models.