| Contents |
| Clinicians are routinely following the evolution of the tumor of a patient thanks to imaging devices. They are evaluating the growth or a response to a treatment using quantitative criteria (like RECIST) computed from these images.
Our goal is to improve these criteria by using mathematical models for tumor growth and to be able to quantify the agressivness of a tumor or even obtain a prognosis. For this matter, two main challenges have to be overcome. First, one has to design an accurate deterministic mathematical model of the disease that is able to reproduce the observed behaviors on patients. These models typically involve many parameters (which are e.g tuning the interplays between the various phenomena influencing the tumor growth) that are practically difficult or even impossible to recover them from experimental data or clinical routine. Hence, the second challenge is to find a way to recover reasonable values of these parameters, that allow the computed evolution to match the observed ones, exclusively with the information available to clinicians.
I will describe these challenges and the methods we have developed to overcome them in the clinical context of evaluating the agressivness of metastases (from bladder, thyroid) to the lung. |
|