EN FR
EN FR


Section: Research Program

Stroke

Project-team positioning

Many teams work on medical imagery, in order to improve diffusion images or MRIs, or to improve their mathematical analysis. But there are few academic teams working on ischemic stroke modeling in Europe, mainly because there are few available clinical data for a given patient. Except routine imagery, often with low definition because it is taken very quickly, in an emergency context, there are no biochemical data. The follow up is also very sparse.

However the domain is very rich. Ischemic stroke involves blood flow, ionic exchanges, cell swelling, cell death (including necrosis and apoptosis), reperfusion, free radicals ... It is possible to build very detailed descriptive models of stroke, however it is not possible to parametrize these models using clinical data. Only a crude parametrization using bibliographical data or ad hoc parameters is possible. We followed this path at the beginning of Numed. Now we are focused on clinical data, namely diffusion and perfusion maps.

The scientific challenge is to try to predict the outcome of the patient, starting from the initial clinical images, obtained when the patient enters the hospital. By outcome, we mean the final size of the dead area, or patient abilities (speech, walk, standing). It is particularly important to try to know whether clinicians have to give some particular drug to try to reopen blood flow, or whether this reopening would be harmful for the patient.

Results

Through the PhD thesis of Mathilde Giacalone, we built a strong collaboration with an imagery team of the Creatis lab (Lyon I). This team is very strong in the analysis of clinical images of stroke and in particular on perfusion. It closely works with clinicians (Pr Nighoghossian at Lyon Bron hospitals). We have therefore built a strong connection with clinicians and image specialists.

Thanks to the work of Mathilde Giacalone, we have now at hand a large database of clinical data (more than 50 patients). For each of these patients we have at hand their perfusion and diffusion maps at their entry, together with diffusion one week and one month later. All these images have been checked and segmented. All the data have been supervised by a clinician. The database is therefore very reliable. Its setup took almost two years. This database will grow with time, according to the patients treated in Lyon.

Mathilde Giacalone worked on new algorithms to improve perfusion images. A perfusion image describes the blood flow in the brain. We also fulfilled a preliminary analysis to know what kind of image (there are several protocols to get perfusion images) is the best to predict the outcome of the patient, using information theory.