Modeling tumor growth: from differential deformable models to growth prediction of tumors detected in PET images

Modeling of a growing tumor over time is extremely difficult, because of the complex biological phenomena underlying cancer proliferation. Existing models can mostly describe in vitro experiments of spherically-shaped avascular tumors, but they cannot match the highly heterogeneous and complex-shaped tumors seen in cancer patients. We propose a new time-dependent geometric deformable model that can characterize tumors of complex shape, such as vascular tumors. Preliminary result show that such a model can provide an adequate framework for analyzing and predicting the growth of tumors seen in PET images of cancer patients.